PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:/Prolecc&Pres/CIRC/DaIeDuerindd

CHANGING GEARS:
MODELING GENDER DIFFERENCES IN PERFORMANCE
ON TESTS OF MECHANICAL COMPREHENSION
By

James A. Grand

A THESIS

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

MASTER OF ARTS
Psychology

2008

ABSTRACT
CHANGING GEARS:
MODELING GENDER DIFFERENCES IN PERFORMANCE
ON TESTS OF MECHANICAL COMPREHENSION
By
James A. Grand
For over 60 years, the selection, training and vocational coaching assessments of
individuals in mechanically-inclined occupations has often involved the administration of
mechanical ability tests. However, men have consistently outperformed women on such
tests, creating the possibility for unfair hiring or human resource practices in these
situations. Unfortunately, the causes of these achievement differences are still largely
unknown. As such, this study proposes and examines a functional gender differences
model that examines predictors of mechanical ability test performance beyond gender
that could potentially be leveraged to diminish the substantial performance differences
found in mechanical and other similarly gender-biased cognitive ability tests.
Speciﬁcally, the model investigates one’s gender role identiﬁcation, gender stereotype
endorsement, mechanical interests/experiences, and mechanical self-efﬁcacy as
inﬂuences of mechanical comprehension. The results of the study revealed that, even
when controlling for gender, an individual’s self—efﬁcacy for accomplishing
mechanically-related activities was strongly and positively related to test performance;
furthermore, self-efﬁcacy was predicted by one’s mechanical interests and experiences,
which were meaningfully related to gender role identiﬁcation. Practical implications and

limitations of the model’s results in terms of capturing gender differences in cognitive

ability tests and topics for future related research are discussed.

TABLE OF CONTENTS

LIST OF TABLES .............................................................................................................. v
LIST OF FIGURES .......................................................................................................... vii
INTRODUCTION .............................................................................................................. 1
Deﬁnition of the Mechanical Comprehension Construct ............................................. 3
Early Research on Mechanical Comprehension ........................................................... 5
Empirically Observed Gender Diﬂerences in Mechanical Comprehension ................. 7
Theories of Gender Diﬁ'erences in Mechanical Comprehension ................................ 11
Proposed Model of Gender Differences in Mechanical Comprehension ................... 16
Hypothesis 1 .............................................................................................................. 20
Hypothesis 2 .............................................................................................................. 22
Hypothesis 3a ............................................................................................................ 28
Hypothesis 3b ............................................................................................................ 28
Hypothesis 4 .............................................................................................................. 35
Hypothesis 5 .............................................................................................................. 44
Hypothesis 6a ............................................................................................................ 44
Hypothesis 6b ............................................................................................................ 45
Hypothesis 6c ............................................................................................................ 45
Hypothesis 6d ............................................................................................................ 45
METHOD ......................................................................................................................... 46
Sample ......................................................................................................................... 46
Measures ..................................................................................................................... 46
Procedure .................................................................................................................... 5 1
RESULTS ......................................................................................................................... 53
Scale Analyses ............................................................................................................. 53
Mechanical Interests, Knowledge, and Experiences ................................................. 54
Mechanical Self-Efﬁcacy .......................................................................................... 58
Gender Role Identiﬁcation ........................................................................................ 58
Gender Stereotype Endorsement ............................................................................... 60
Mechanical Comprehension ...................................................................................... 62
Hypothesis Tests .......................................................................................................... 64
Hypothesis 1 ......................................................... 64
Hypothesis 2 .............................................................................................................. 66
Hypothesis 3a ............................................................................................................ 69
Hypothesis 3b ............................................................................................................ 69
Hypothesis 4 .............................................................................................................. 70
Hypothesis 5 .............................................................................................................. 72
Hypothesis 6a ............................................................................................................ 75
Hypothesis 6b ............................................................................................................ 75

iii

Hypothesis 6c ............................................................................................................ 75

Hypothesis 6d ............................................................................................................ 75
Additional Analyses .................................................................................................... 78
Between-Gender Tests .............................................................................................. 78
Male-Female Stereotype Endorsement ..................................................................... 81
Hypothesis 5 Follow-up ............................................................................................ 85
DISCUSSION ................................................................................................................... 88
The Gender Difference Model and Specific Ability Testing ....................................... 90
Gender-related characteristics ................................................................................... 9O
Contextually-related characteristics ........................................................................ 1 13
Capturing Male-Female Differences: An Investment Well-Spent ............................ 118
Limitations ................................................................................................................ 120
Conclusion ................................................................................................................ 123
APPENDIX A ................................................................................................................. 124
APPENDIX B ................................................................................................................. 125
APPENDIX C ................................................................................................................. 127
APPENDIX D ................................................................................................................. 128
APPENDIX E ................................................................................................................. 129
APPENDIX F .................................................................................................................. 130
APPENDIX G ................................................................................................................. 13 1
APPENDIX H ................................................................................................................. 132
REFERENCES ............................................................................................................... 135

iv

LIST OF TABLES

Table 1. Cognitive Gender Diﬂerences (ds) on the Mechanical Reasoning Subtest of the
DAT by Grade and Year of Standardization ....................................................... 10

Table 2. Percentage of Subjects in Bem ’s (1974) Five Gender Role Categories for each of
the Four Gender Roles Deﬁned by Spence, Helmreich and Stapp (I 975) ......... 24

Table 3. Mean Spatial Reasoning Performance (SD) on the Diﬂerential Aptitudes Space
Relations Test as a Function of Gender and Gender Preference ....................... 38

Table 4. Mechanical-Reasoning Means Based on Median Split Method Categories ....... 39
Table 5. Demographic Information of Study Sample (N = 258) ...................................... 47

Table 6. Unrotated and Rotated (Varimax) Factor Loading Matrix for Items from the
Mechanical Interests, Knowledge and Experiences Subscales ........................... 56

Table 7. Items, Reliabilities, Means and Standard Deviations (SD) for Mechanical
Experience and Mechanical Interest Scales ....................................................... 59

Table 8. Unrotated and Rotated (Varimax) Factor Loading Matrix for Items from the
Gender Stereotype Endorsement Scale ............................................................... 61

Table 9. Items, Reliabilities, Means and Standard Deviations (SD) for Female Stereotype
and Male Stereotype Endorsement Scales .......................................................... 63

Table 10. Means, Standard Deviations and Correlations of Study Variables .................. 65

Table 11. Means, Standard Deviations, and Effect Sizes of Study Variables Across
Gender ............................................................................................................... 66

Table 12. Regression Coefﬁcients for Mechanical Interest and Experience in Predicting
Mechanical Self-Eﬂicacy (Hypothesis 1) .......................................................... 66

Table 13. T est for Mediating Eﬂects of Mechanical Self-Eﬂicacy on Relationship Between
Mechanical Interest and BMC T Scores (Hypothesis 2) .................................... 67

Table 14. Eﬂects of Gender Role Identiﬁcation and Gender Stereotype Endorsement on
Mechanical Self-Eﬁ‘icacy (Hypothesis 4) .......................................................... 71

Table 15. Eﬂects of Gender, Gender Role Identification and Gender Stereotype
Endorsement on BMC T scores (Hypothesis 5) ................................................. 73

Table 16. Effects of Gender and Gender Role Identification on BMC T Performance
(Hypotheses 6a — 6d) ......................................................................................... 77

Table 17. Regression Coeﬂicients for Mechanical Interest and Experience in Predicting
Mechanical Self-Efficacy Beyond Gender (Hypothesis 1) ................................ 79

Table 18. Mediating Effects of Mechanical Self-Efficacy on Relationship Between
Mechanical Interest and Experience and BMC T Scores Controlling for Gender

(Hypothesis 2) ................................................................................................... 80
Table 19. Eﬂects of Gender Role Identification and Gender Stereotype Endorsement on

Mechanical Self-Efficacy (Hypothesis 4) .......................................................... 82
Table 20. Hypothesis Summary ......................................................................................... 89
Table 21. Rotated (Varimax) Factor Loading Matrix for Items from the BSRI ............... 99

Table 22. Mean BMC T Scores Based on Median-Split, BSRI Gender Role Identification
Categories (Hypotheses 6a — 6d) .................................................................... 133

vi

LIST OF FIGURES

Figure 1. Diagram of hierarchical structure of human abilities (Vernon, 1950). .............. 4

Figure 2. Proposed model of gender diﬂerences in mechanical comprehension as
measured by test performance. .......................................................................... 17

Figure 3. Example of the hypothesized effects of gender role identiﬁcation on an
individual ’s level of mechanical comprehension. .............................................. 27

Figure 4. Predicted interaction effect of gender role identification and gender stereotype
endorsement on reported levels of mechanical self-eﬂicacy. ............................ 36

Figure 5. Predicted three-way interaction eﬂect of gender, gender role identification and
gender stereotype endorsement on test performance ......................................... 43

Figure 6. Example items from BMC T F orm S ................................................................... 50

Figure 7. T wo-way interaction between female stereotype endorsement and male
stereotype endorsement on BMC T performance for females and males ........... 84

Figure 8. T wo-way interaction between feminine gender role identiﬁcation and female
stereotype endorsement on BMC T performance for females and males ........... 87

vii

CHANGING GEARS:
MODELING GENDER DIFFERENCES IN PERFORMANCE
ON TESTS OF MECHANICAL COMPREHENSION
INTRODUCTION

1 was born a mechanic, and made a barrel before I was ten years old. The
cooper told my father, “Fanny made that barrel, and has done it quicker and
better than any boy I have had after six months’ training. ” My father looked at it
and said, “What a pity that you were not born a boy so that you could be good
for something. Run into the house, child, and go to knitting. ”

—— Frances D. Gage (Stanton, Anthony, & Gage, 1882)

The opening quotation above from American suffragist Frances Gage succinctly
and strikingly highlights an issue that has long stirred debate in court rooms, board rooms,
class rooms and living rooms the world over—are men and women inherently different in
terms of what they can and can not do or are such gender differences only a reﬂection of
the environment in which we live? Such a deﬁnitive and encompassing answer to this
question is not to be found in the current research, nor is such a resolution likely to be
found in any research in the near-future as well. But, by limiting the scope of broad
gender differences in aptitude to a more narrow range of dimensions, it may be possible
to incrementally build a better understanding of the processes through which such
differences arise and are maintained. As such, the present study’s focus is concerned
with understanding the differences observed between males and females in the speciﬁc
cognitive ability of mechanical comprehension.

Though much more will be relayed on this topic in the sections to follow,
empirical research has long noted the superiority of males over females on cognitive tests
of mechanical comprehension (e. g., Bennett & Cruikshank, 1942). This ﬁnding is
particularly troubling given that such tests are used for a wide variety of industry

practices in mechanically related jobs, including selection, training assessment and

vocational counseling (Super & Crites, 1962). The severity of the problem is ampliﬁed
even further when the status of the current labor market is brought into consideration.
Within the last twenty years, the number of women who have begun to move into
traditionally male dominated work has increased dramatically (Blau & Hendricks, 1979;
England, 1981; Beller, 1982), including jobs for which tests of mechanical
comprehension are typically administered as a means of selection (Super & Crites, 1962).
Though such mechanical jobs (such as production and maintenance work) are seldom
considered “high-status,” they often offer better pay and more steady employment than
comparative female dominated occupations (such as clerical work, Muchinsky, 2004).
Thus if females truly are deﬁcient in mechanical comprehension, administering tests for
this construct as part of a selection system will almost certainly prevent women’s entry
into these potentially more attractive mechanical jobs; however, if the observed male-
female difference is instead a response to some sociocultural phenomenon (i.e.,
stereotypes, gender roles, etc.), it may be possible to reduce the performance gap through
training or some other means of education.

The purpose of this research is to propose. and test a model of the antecedents
related to the performance disparity across gender found on tests of mechanical
comprehension (also referred to as mechanical aptitude or ability). The development of
the components and linkages of the model will be established through an overview of
previous research in the areas of mechanical comprehension, gender differences, and test
performance. Following this review, the study’s research hypotheses are given and
discussed in relation to the gender model and the relevant literature, and the

methodological procedures undertaken to test these hypotheses are described. Finally, an

analysis of the data is presented and interpretations of the study’s conclusions are drawn
for the purposes of aiding future research and development in the area of gender
differences in mechanical ability testing speciﬁcally, and hopefully cognitive ability

testing in general.

Deﬁnition of the Mechanical Comprehension Construct

It is perhaps helpful to approach the construct of mechanical comprehension
through the framework of Spearman’s (1927) Two-Factor theory of intelligence.
According to the theory, a person’s cognitive functioning can be broken down and
analyzed into two distinct factors—a general (g) factor and speciﬁc (s) factors. As the
name implies, g is an overall mental capacity or intellectual capability that operates in all
forms of cognitive functioning. s is similarly deﬁned, but rather than operating across
many areas of cognitive functioning, it is more narrowly restricted in its range. With the
development of more sophisticated factor analysis techniques (Guilford, 1948; Thurstone,
1948), researchers soon after modiﬁed the Two-Factor theory of intelligence to include
lower- and hi gher-order factors that lie in between g and s in terms of their inﬂuence on
the performance of certain abilities.

Vernon (1950) combined these ﬁndings to create the hierarchical structure of

human abilities, which demonstrated how g could be usefully conceptualized as the
hi ghest-order factor among more speciﬁc levels of intellectual functioning. The model
extended from the various permutations of s on up through successively more
encompassing factors (see Figure 1). Through factor analysis of g ability tests

administered in selective military populations, Vernon discovered that two major group

 

 

 

 

 

Major group factors l 1
vse k:m
Minor group factors I . l
A A I —. A r——. I A
Specmcfactors I||||||llllllllllllllllllllllll||||||||||||||||||lll|

Figure l . Diagram of hierarchical structure of human abilities (Vernon, 1950).

factors reliably emerged. The ﬁrst of these he labeled a verbal-numerical-educational
factor (represented as vsed), and the other a practical-mechanical-spatial-physical factor
(ksm). Given sufﬁciently speciﬁc tests, Vernon states, each of these types further
subdivides into a series of minor group factors. In the case of the ksm factor, these
include spatial, manual and mechanical information subfactors, among others. Based on
Vernon’s hierarchical structure, then, it can be concluded that mechanical comprehension
does indeed appear to load on both g and spatial/perceptual factors (Guilford, 1947).
Indeed, empirical work seems to support this theory as Bennett (1969) reports that early
versions of the Bennett Mechanical Comprehension Test (BMCT) demonstrated
correlations between .40 and .60 with various intelligence and spatial ability tests.
Nevertheless, because of the high face validity, reliability, and predictive validity for tests
of mechanical comprehension among jobs requiring the aptitude, the construct has gained
considerable support as one of the most accurate cognitive assessments of performance in
mechanically-inclined occupations (Muchinsky, 2004).

As demonstrated by Vernon (1950), a number of cognitive abilities are closely
related to mechanical comprehension; thus it is necessary to precisely identify the

construct for the purposes of the current research study. To this end, mechanical

comprehension will herein be deﬁned as the ability to learn about, perceive and
understand the operation of common physical principles and mechanical elements in
practical situations, in both an implicit and explicit manner (Bennett, 1969; Weisen,
1999). This deﬁnition was speciﬁcally chosen to emphasize the natural, instinctive and
intuitive fashion through which mechanical comprehension is experienced, as opposed to
any proﬁciency gained through formal instruction in physics or mechanical diagnostics.
In addition, it is intended to exclude any other closely related cognitive factors, such as
spatial visualization (Wittenbom, 1945; Guilford, 1948) or perceptual speed and acuity

(Super & Crites, 1962).

Early Research on Mechanical Comprehension

Spurred by the widespread acceptance of cognitive ability testing during the post-
World War I/pre-World War II era, researchers began to take a vested interest in the
study of mechanical aptitude in the early 19003. Although tests of mechanical aptitude
had been in circulation for nearly half a decade (Stenquist, 1923), the ﬁrst empirical
studies of mechanical comprehension were not published until 1928 by Cox and, shortly
thereafter, Patterson, Elliot, Anderson, Toops, and Heidbrieder in 1930.

Following Spearrnan’s (1927) Two-Factor theory of intelligence, Cox (1928) set
out to determine whether mechanical aptitude was a distinguishable construct apart from
g. Through a series of paper-and-pencil tests, Cox attempted to isolate and capture the
unique cognitive mechanisms involved in mechanical comprehension. In order to
minimize any potential construct contamination, the tests were designed such that any
instruction in physics or formal mechanics was not necessary or speciﬁcally useful in

answering the questions. For example, the Test of Mechanical Models presented

participants with the beginning and end states of a particular mechanical operation from
which they were to derive the intermediary processes. Based on factor analysis of his
data and a number of divergent validity studies, Cox concluded that there was indeed a
cognitive ability distinct from g that predicted performance on tests of mechanical
comprehension which he termed mechanical aptitude. Although much later, researchers
have suggested that Cox actually captured the closely related spatial visualization factor
rather than a distinctive mechanical comprehension dimension in this initial study (Super
& Crites, 1962); nevertheless, these efforts generated a great deal of early empirical
interest in the area of mechanical aptitude.

Unlike the strictly theoretical approach adopted by Cox (1928), Patterson et a1.
(1930) set out to develop a practical measure that could be used to identify individuals
who displayed a proﬁciency in mechanical comprehension. According to Patterson et
al.’s deﬁnition, mechanical ability is taken “to refer to whatever capacities and abilities
are necessary for certain kinds of work—speciﬁcally, work that involves the
manipulation of tools, the operation of machinery, and the planning and execution of
pieces of work which involve these and similar activities” (p. 5). Rather than create a
unique bank of tests for these purposes, the researchers revised and selected items from a
number of previously published tests including the Minnesota Assembly test, Minnesota
Spatial Relations test, and Minnesota Paper Form Board test, among others. Based on an
extensive criterion validity effort, a weighted average scoring system was derived to
achieve a ﬁnal, unique score for mechanical ability. Patterson et a1. administered their

ﬁnal test battery to a large number of junior high school boys, in the process establishing

a large normative dataset, which would serve as the standard for future tests of
mechanical comprehension.

In the years following Cox (1928) and Patterson et al.’s (1930) inquiries, tests of
mechanical comprehension were readily adopted into industry practice for a variety of
purposes, including admission into training programs (Martin, 1951) and trade schools
(Patterson, 1956), vocational counseling (Cantoni, 1955; Klare, Gustafson, Mabry &
Shuford, 1955), and, especially, as instruments of selection (Harrell & Faubion, 1940;
Ghiselli & Brown, 1951; Wolff & North, 1951). In addition, mechanical aptitude testing
has been applied extensively in the military arena for over 60 years as a means of
enlistment and duty assignment for identifying individuals qualiﬁed for service in a
particular branch. Even as early as 1947, Guilford and Lacey found that the BMCT—
which to this day remains the most widely used test of mechanical comprehension
(Muchinsky, 2004)——was one of the top predictors of performance by military ﬁghter
pilots during World War II. In fact, tests of mechanical comprehension were widely used
during World War 11, so much so tha “...the War Department considered the
measurement of mechanical ability of such importance that a Mechanical Aptitude test
was administered at reception centers to all enlisted men who could read or write” (Tiﬁ'rn,
Knight & Asher, 1946, p. 236). It would appear, then, that the construct of mechanical
comprehension has had a very real impact from both an empirical and practical

perspective.

Empiricallv Observed Gender Dﬂferences in Mechanical Comprehension
The measurement and identiﬁcation of mechanical comprehension has had a

fairly long history of research and industry use—however, very few advances in terms of

our understanding of the construct and our ability to accurately assess it have been made
in recent years (Stumpf, 1995). Muchinsky (2004) suggests that this may be a result of
the “‘unfashionable’ nature of studying industrial production workers” (p. 23) or perhaps
because of its relatively restricted area of applicability in today’s workforce. Whatever
the case may be, the lack of progress is unfortunate, especially given that mechanical
aptitude is one of the most reliable and consistent cognitive differences across genders
(Bennett & Cruikshank, 1942; Antill & Cunningham, 1982; Anastasi, 1988; Stumpf,
1995; Halpern, 2000).

A number of studies have observed the marked gender difference in performance
on tests of mechanical comprehension. In assessing a variety of standardized aptitude
tests, Stanley, Benbow, Brody, Dauber, and Lupkowski (1992) noted signiﬁcant gender
effect sizes on tests of mechanical ability, with d coefﬁcients ranging from .66 to .89.
Bennett and Cruikshank (1942) also reported large gender differences in performance
when conducting validation studies on an earlier version of the BMCT. An
approximately 15 point mean difference in test scores was observed between males and
females in Grades 10 through 13; using somewhat older age groups (applicants to nursing
schools, M = 23.37, versus applicants for positions as ﬁremen and policemen, M = 36.78)
resulted in only a very slight improvement in the performance gap. Not surprisingly, the
authors found gender to be a signiﬁcant predictor of performance on the test, explaining
from 49% to 79% of the variance in test scores across the different samples.

A somewhat more evolved pattern of gender differences in mechanical
comprehension test performance was found using extensive validation data collected for

the Mechanical Reasoning subscale of the Differential Aptitudes Test (DAT). Bennett

(1969) reported that while the mean within-group test scores for both males and females
increased each year between Grades 8 and 12, the mean difference between groups (i.e.
the performance gap) also increased each year. Thus, although males and females
demonstrated improved scores on the test from one year to the next, the improvement by
males outpaced females such that the difference in performance was actually much
greater at older ages.

Feingold (1988) further expounded these results by analyzing approximately 30
years of normative data collected on the eight subtests of the DAT for gender differences
in test performance. The portion of Feingold’s results concerning mechanical
comprehension is reproduced in Table 1. As can be seen from the table, males once
again signiﬁcantly outperformed females on the Mechanical Reasoning subtest at every
grade level and at every year. Though not apparent from the limited data shown in Table
1, the'magnitude of the gender differences in performance on Mechanical Reasoning
were actually the largest of any of the eight DAT subscales. The analysis by Feingold
also revealed a signiﬁcant age effect, such that the performance gap on tests of
mechanical comprehension between males and females grew larger at each successive
age group. The ﬁndings by Bennett (1969) and Feingold (1988) that the performance gap
on tests of mechanical comprehension widens as a function of age is a common one in the
literature on developmental gender differences in speciﬁc cognitive abilities (Maccoby &
Jacklin, 1974); however, mechanical aptitude appears to be one of the most intensely
pronounced (Anastasi, 1981; 1988).

As an aside, the mechanical comprehension data from the Feingold (1988) study

also demonstrates an interesting “secular trend” that has been noted by many researchers

Table 1
Cognitive Gender Diﬂerences (ds) on the Mechanical Reasoning Subtest of the DAT
by Grade and Year of Standardization

 

 

 

 

Grade in School
Year 8 9 10 11 12 Y)?!”
1947 1.02 1.24 1.34 1.48 1.55 1.33
1962 .80 .90 1.00 1.06 1.23 1.00
1972 .74 .79 .84 .88 .92 .83
1980 .66 .68 .75 .82 .89 .76
Grade M .80 .90 .98 1.06 1.15 ,9_8

 

Positive d indicates higher male performance.
Table reproduced from F eingold, A. (1 98 8). Cognitive gender differences are disappearing.
American Psychologist, 43(2), 95-103.

studying longitudinal gender differences in speciﬁc cognitive ability (Stumpf, 1995).
The secular trend refers to the shrinking performance gap in speciﬁc cognitive abilities
between male-female cohorts (represented by examining the columns in Table 1 ﬁom top
to bottom), as opposed to the increasing performance gap within male-female cohorts
(represented by examining the rows in Table 1 ﬁ'om left to right). In large part, there is
empirical uncertainty as to why such a pattern has been observed. Flynn (1984) reported
over two decades ago that the average American’s IQ (a general measure of g) increased
by 12 points between 1932 and 1972, thus hypothesizing that there may be a generational
effect on cognitive ability (however, this “Flynn Effect” has not been universally
replicated, see Sundet, Barlaug, & Torjussen, 2004; additionally, there is at least some
evidence stating that the growth has slowed in recent years, Teasdale & Owen, 2000).
Extrapolating from this hypothesis, the hierarchical structure of human abilities
(Vernon, 1950) suggests that an increase in g should positively correlate with increases in
particular speciﬁc cognitive abilities as well. Therefore, it could be argued that gains in g

for women have contributed to greater improvements in mechanical ability while

10

manifesting in different areas of functioning for men. At a more universal level of
explanation, it is likely that temporally—related changes in the brain/biology and
sociological characteristics (see section below on Theories of Gender Diﬂerences) of the
general population have contributed signiﬁcantly to the observed secular trend as well
(Teasdale & Owen, 2000). Of course, these speculations are hypothetical, and the causes
and explanations for this trend are still unclear in the research literature; however, despite
the shrinking performance gap in mechanical comprehension, the magnitude of the male-

female differences that do still exist are quite large and thus warrant ﬁirther study.

Theories of Gender Differences in Mechanical Comprehension

At the deﬁnite risk of oversimpliﬁcation, there are two broad classes of
explanations that attempt to capture how and why gender differences in aptitudes arise:
biological factors and direct socialization inﬂuences from the environment (Maccoby &
Jacklin, 1974, present an extensive review of much of the early research conducted in
these areas and serves as an excellent point of entry into the discussion). However, it is
important to note that these theories are not mutually exclusive, and they should not be
considered as such. It is simply convenient to separate and study both inﬂuences as
though they were distinct in order to better isolate the underlying phenomena captured by
each. There is little argument that the interaction between nature, nurture and the various
psychological processes that accompany them is the most accurate description of the
underlying differences that exist between men and women (Wood & Eagly, 2002). But,
as is often the case, the requirements and limitations of empirical research outweigh our
ability to wholly capture the true nature of the event and thus we must rely on a less

desirable, “piecemeal” approach. Nevertheless, a great deal of research effort has gone

11

into explaining the origins and sustaining forces of gender differences in speciﬁc
cognitive abilities, of which only a cursory glimpse will be presented below.

Biological factors. The argument for biological differences between the genders
accounting for differential cognitive performance has been approached in a number of
different ways and has implicated a number of possibilities. One of the most widely
studied areas in this arena concerns the effect of hormones on cognitive ability (e. g.,
Rogers, 1976; Messent, 1976; Harris, 1978; 1981). The basic notion follows that during
crucial brain development stages, males and females are exposed to and come to possess
differing amounts of the primary sex hormones, including estrogen, progesterone and
testosterone, among others (Hamburg & Lunde, 1966; Halpern, 2000). Based on the
relative proportion and activation of these hormones, areas of the brain develop
differently—thus, as there should typically be less within-sex variability than between-
sex variability in terms of relative levels of the sex hormones, the brains of males and
females should tend to develop in different fashions thereby resulting in structural brain
differences across the sexes that could account for variability in performance on certain
tasks.

For example, one of the most inﬂuential theories concerning the impact of
hormones on cognitive ability was put forth by Geschwind and Galaburda (1987). In
this theory, it was proposed that higher concentrations of prenatal testosterone slow the
growth of neuronal connections within the brain’s left hemisphere. This often results in
right hemisphere dominance (also known as brain lateralization), in which the right
hemisphere of the brain has much more ﬁnely tuned control and “processing power” than

the leﬁ hemisphere. As a normally developing male will be exposed to higher levels of

12

prenatal testosterone than a normally developing female (because testosterone is naturally
produced by the male testes), males are much more likely to be right-brain dominant, a
ﬁnding which has been reliably reproduced by many researchers (cf. Halpem, 2000). On
the other hand, females are more likely to exhibit bilateral brain functioning (equal
activation in both hemispheres) and are, consequentially, much more likely to employ
both hemispheres when approaching a cognitive task (Harris, 1981).

Thus, Geschwind and Galaburda’s (1987) theory predicts that males should be
more efﬁcient at tasks that require primarily right-brain activation while being less
proﬁcient at left-brain tasks—and indeed, there has been some research to support this
claim. Mathematical reasoning and spatial ability skills represent two notable cognitive
abilities that have been demonstrated to be right-brain dependent, and two notable areas
in which males have typically outperformed females (Maccoby & Jacklin, 1974; Harris,
1981; Halpem, 2000). In addition, males have also been found to possess the majority of
the language production and reading problems (such as dyslexia) normally associated
with active left-brain functioning. Furthermore, these cognitive verbal abilities are both
areas in which females have demonstrated markedly better performance than males
(Maccoby & J acklin, 1974; Halpem, 2000). Based on these ﬁndings then, and recalling
the earlier discussion of the relationship between mechanical comprehension and spatial
ability (Vernon, 1950), there is at least indirect evidence to support the notion that the
differences in performance found on tests of mechanical comprehension can partly be
attributed to the differential brain functioning between males and females.

Gender socialization. The quotation from Frances Gage presented at the

beginning of this paper provides a blunt, yet poignant, example of the gender

13

socialization process. Theories of gender socialization generally posit that, over the
course of the natural lifespan and especially during the inﬂuential childhood years, males
and females are differentially treated by others in their immediate environment as a result
of established sociocultural norms. Such treatment manifests in the disparate types of
opportunities, reinforcement/punishment, and training that are made available to the
members of each sex. As a result of these experiential differences then, males and
females come to acquire and gain proﬁciency in dissimilar sets of skills and aptitudes
(Maccoby & Jacklin, 1974; Jacklin & Reynolds, 1993; Lott & Maluso, 1993; Wood &
Eagly, 2002). Perhaps put more succinctly, gender socialization is the process by which
individuals come to learn and adopt the socially created and accepted norms, goals and
values of their gender (Eccles, 1987). Since these gender-appropriate goals and values
are typically quite different, men and women oﬂen become better or more skilled with
certain tasks and abilities than members of the opposite sex (Stumpf, 1995; Halpem,
2000).

There are a number of different theories that overlap with or fall under the
umbrella of “gender socialization.” For example, social learning theory (Bandura, 1969),
though typically treated as a more general explanation for the behavioral adoption
patterns seen in children, is very similar to the above treatment of gender socialization.
In social learning theory, children are reinforced for certain desired behaviors by the
social agents in their environment. In addition, social learning theory also emphasizes
the importance of behavior modeling, in which children choose to imitate the behaviors
of the inﬂuential others around them. From the perspective of gender then, social

learning theory predicts that children choose to imitate models of the same sex, and thus

14

acquire the behaviors and associated skills and abilities displayed by their male or female
role models.

Also similar to the gender socialization hypothesis is gender schema theory,
which proposes that individuals develop a systematic, formulated set of ideas about what
is and what is not male or female based on the information available to them in the
environment (Bern, 1981). This network of associations (which incorporates a diverse
number of categories including perceptions about affect, attitudes, objects, behaviors, etc.)
then becomes a “cognitive ﬁlter” through which any new information and experiences
pertaining to men and women are passed through, organized, and then adapted into the
previous gender schema. To this end, one’s gender schema is constantly changing and
evolving to incorporate and process new information pertaining to the roles of males and
females. One of the more important implications of gender schema theory is that
individuals learn to “sort people, behavior, and attributes into the culture’s deﬁnitions of
feminine and masculine” (Jacklin & Reynolds, 1993, p. 201). It logically follows, then,
that individuals would also attempt to place themselves within their gender schema based
on their own behaviors, attitudes, etc. Therefore, those people who are “gender
schematic” and adhere rigidly to their associations pertaining to gender roles are more
likely to adhere to the sociocultural norms about male-female differences in skills and
aptitudes than individuals who are “gender aschematic” and not as reliant on gender-role
associations (Bern, 1985).

In terms of the impact of gender socialization on the development of mechanical
comprehension, a return to Frances Gage’s short-lived experience as a mechanic’s

apprentice depicted earlier yet again offers an exceptionally concrete and vivid means of

15

explanation. Quite clearly, Gage believes and deﬁnitively states that she has always had
some innate proﬁciency with all things mechanical, and has even demonstrated that she
can perform a mechanical task as well or better than any similarly trained male. However,
even despite her obvious aptitude, her father wholly discourages the pursuit of such a
traditionally masculine activity and instead instructs her to return to the house and take up
a more “appropriate” feminine activity. Presumably, once “in the home” and in the
presence of other same sex feminine social agents, Gage is not likely to ﬁnd an inﬂuential
model that she could imitate or learn from in order to build the physical skills and
cognitive abilities required of a mechanic. Thus, she is denied the opportunity to pursue
further training and is shut out from additional experiences that might have served to

improve her level of mechanical comprehension because of a sociocultural norm.

Proposed Model of Gender Differences in Mechanical Comprehension

The functional model of gender differences in mechanical comprehension that
will be tested by the current study is presented in Figure 2. Though the major variables
and hypotheses concerning their linkage will be discussed and justiﬁed in the sections to
follow, a general overview of the model serves as a useful starting point. The basic
organization follows a direct mediation structure in which self-efﬁcacy for mechanical
tasks is proposed to inﬂuence the degree to which an individual’s level of mechanical
interest, knowledge, and experience predicts performance on a test of mechanical
comprehension. Two additional variables are also introduced which are believed to have
an impact on the direct mediating linkage. The ﬁrst, gender role identiﬁcation, is

predicted to inﬂuence the level of mechanical interest, knowledge, and experience

16

.coceszotca 88 an 8882: 8 8688888 80338:. 5 80:8ch 828w .8 .385 @8395 .N Miami

 

 

 

 

 

 

 

 

 

 

 

 

 

 

«8:88.35 8585882
2.5883. .8380 8.3. .8980
4
88888.25 >uqum 82.2896
82:28.2 .0 r 4 98 828.265.

 

88... co coca—Eaten. Eem .838:er .8888. 83888.2

 

 

 

 

 

 

 

 

17

reported by an individual. The second, endorsement of gender stereotypes, is predicted to
moderate the link between mechanical interest, knowledge, and experience and
mechanical self-efficacy. The overall model can be observed for both males and females
to identify where inconsistencies in the linkages lie.

The model proposed by the current study offers a signiﬁcant point of departure
from previous explanations of gender differences in speciﬁc cognitive ability in that it
focuses more attention on the individual-level, process-outcome links that result in
differential aptitude testing scores. As the previous section on theories of gender
differences outlined, a great deal of theoretical and empirical effort has gone towards
identifying the broad antecedents that account for group-level gender differences—but far
less consideration has been given to the interaction between gender differences and
individual differences (i.e. mechanical experiences, mechanical self-efﬁcacy, etc.) that
might impact the measurement of speciﬁc aptitudes. The fact that gender plays an
important role in the development and expression of certain cognitive abilities is not
questioned by the current study; however, the direct gender-to-performance relationship
seems an overly simpliﬁed explanation. Though the link that connects gender to aptitude
testing performance has been well documented, there is still a signiﬁcant amount of
variance left to be explained (Antil & Cunningham, 1982). Understanding potential
individual difference variables that contribute to the performance gap and how they
interact would help to more precisely elucidate how variability in speciﬁc aptitudes plays
out across gender and gender-related characteristics. Thus one of the goals of the present

research is to validate a relational model that incorporates known outcomes of gender

18

differences with relevant individual differences in order to explain performance variation
on tests of mechanical comprehension.

Mechanical interests, knowledge, and experiences. Bandura’s (1977) early social
cognitive theory proposed that feelings of self-efﬁcacy arose primarily from four distinct
sources: performance accomplishments (mastery), vicarious experience (modeling),
verbal persuasion, and emotional arousal. While conceptually useful, these categories are
fairly broad and not always helpﬁil for identifying their direct inﬂuence on self-efﬁcacy.
To that end, Gist and Mitchell (1992) proposed, though never empirically tested, a
restructuring of Bandura’s social learning theory (1977) in which the determinants of
self-efficacy were thought to lie in a 2 x 2 matrix composed of a dimension’s variability
within an individual and a person’s “controllability of the causal inﬂuence” (p. 196). The
variability of a dimension is conceptualized as high or low, depending on how stable the
determinant is over time and situation, whereas the locus of the determinant can be either
external or internal, based on whether the individual has control over the expression of
the causal inﬂuence. As the external cues are proposed to be entirely a ﬁinction of the
task an individual is faced with (e.g., available resources, feedback, environmental
distractions), they are less important for the purposes of the current study as they should
be a constant across all participants. What should be less consistent, however, are the
low variability/intemal and high variability/intemal determinants of self-efﬁcacy. These
include variables such as knowledge, skills, abilities, and interests, among others.
Though no particular mention is made by Gist and Mitchell (1992) as to the degree of
inﬂuence each of these variables has on self-efﬁcacy, it should nevertheless be expected

that that the direction of the relationship will be positive, such that:

19

Hypothesis 1: Individuals with more interest in, greater knowledge of, and more
experiences with mechanically-related subject matter will report higher ratings of

mechanical self-efﬁcacy.

Mechanical self-eﬂicacy. Self-efﬁcacy can broadly be deﬁned as “...peop1e’s
judgments of their capabilities to organize and execute courses of action required to attain
designated types of performances.” (Bandura, 1986, p. 391). Based on the above, then, it
follows that an individual’s level of self-efﬁcacy should be highly predictive of a targeted
performance outcome—and indeed, self-efﬁcacy has been demonstrated to be a reliable
predictor of performance across many different areas of functioning (e. g. Stajkovic &
Luthans, 1988; Mathieu, Martineau & Tannenbaum, 1993; Moritz, Feltz, Fahrbach &
Mack, 2000); however, the degree to which the relationship is meaningful depends in
large part on the level of speciﬁcity at which it is measured (Chen, Gully & Eden, 2001 ).

In the manner in which Bandura (1977; 1986) describes the construct, self-
efﬁcacy is intended to be restricted to task-, domain- and situation-speciﬁc instances.
Thus, through performance accomplishments, vicarious experiences, verbal persuasion or
emotional arousal directed towards a particular area of ﬁmctioning, an individual’s self-
efﬁcacy for performing domain-related tasks is believed to improve and ultimately
translate into improvements in actual performance. However, a number of researchers
have since pushed for the inclusion of general self-efﬁcacy (GSE)—a singular, “trait-
like” evaluation of one’s capability to perform across all areas—as an adequate predictor
of performance as well (e. g. Sherer et al., 1982; Tipton & Worthington, 1984; Woodruff

& Cashman, 1993; Chen, Gully, Whiteman & Kilcullen, 2000; Chen, Gully & Eden,

20

2001; Schwoerer, May, Hollensbe & Mencl, 2005). They hypothesize that, given enough
exposure to successes in multiple speciﬁc domains, a person can develop a broader
concept of self-efficacy that is applicable in a more global sense to a wide variety of
behavioral outcomes.

Numerous studies have been published in recent years that observe the predictive
ability of SSE and GSE in relation to performance. In large part, the evidence for the
SSE-performance link has been surprisingly consistent across a number of different
domains. A meta-analysis by Stajkovic and Luthans (1998) observed a weighted average
correlation of .38 between SSE and work performance measures; a separate meta-analysis
by Moritz et al. (2000) also noted an average correlation of .3 8, this time correlating SSE
and performance in sport. Various other empirical studies have observed SSE-
perforrnance correlations of .37 (Mathieu, Martineau & Tannenbaum, 1993), .38 (Phillips
& Gully, 1997), and .34 (Lane & Lane, 2001).

Compared to the previous ﬁndings, correlations between GSE and performance
have been signiﬁcantly less deﬁnitive. In one of the few meta-analyses available for this
particular relationship, Judge and Bono (2001) reported an average corrected correlation
of .23 between GSE and job performance, substantially smaller than the SSE-
perfonnance link described previously. Still other singular studies have failed to even
ﬁnd a single signiﬁcant correlation. For example, Chen et al. (2001) found no
relationship between students’ GSE prior to taking an exam and subsequent performance
(though measures of GSE after receiving exam scores did reveal signiﬁcant correlations,
reiterating the importance of performance accomplishments on self-efficacy development,

Bandura, 1977); a similar ﬁnding was also reported by Chen et al. (2000).

21

Based on the evidence reviewed here, a measure of SSE seems the more
appropriate approach for the purposes of the current study. Not only has it been shown to
be a more reliable and potent predictor of performance, it is also relevant to the
speciﬁcity of the remaining model variables and thus maintains a meaningful
methodological consistency (Bandura, 1986; Stajkovic & Luthans, 1988; Chen, Gully
and Eden, 2001). The SSE domain of interest for the present research, mechanical self-
efﬁcacy, will herein be deﬁned as an individual’s judgments of his/her capability to learn
about, perceive and understand the operation of common physical principles and
mechanical elements. Thus, the level of mechanical interest, knowledge and experience
possessed by an individual should be more closely related to mechanical self-efﬁcacy,
which should ultimately be more predictive of performance on a mechanical

comprehension test. Therefore, it is hypothesized that:

Hypothesis 2: The relationship between mechanical interests, knowledge and
experiences and performance on a test of mechanical comprehension will be

mediated by mechanical self-efﬁcacy.

Gender role identification. Broadly categorized, gender roles refer to socially
constructed norms that describe the “psychological characteristics that equip [a person]
for the tasks that their sex typically performs” (Wood & Eagly, 2002, p. 701 ). Gender
roles have traditionally been depicted along a single, bipolar dimension ranging from
masculinity on one end and femininity on the other; however, researchers have long since
broadened this perspective to allow for the possibility that an individual may possess

traits typical of both sexes (Constantinople, 1973). In her seminal article, Bern (1974)

22

presented the Bern Sex Role Inventory (BSRI), which contained two separate subscales
for assessing both masculinity and femininity within an individual, thus suggesting that a
person may possess the psychological characteristics of both males and females. In
practice however, the BSRI still places gender roles on a single continuum, categorizing
individuals as feminine, near-feminine, androgynous, near-masculine, and masculine.

Others have attempted to move even further from the typical dichotomous
continuum. Spence, Helmreich and Stapp (1975) developed a measure similar to the
BSRI, but used median splits on the separate masculinity and femininity scales in an
attempt to develop a more two-dimensional model. This technique resulted in four
unique gender roles: masculine (high masculine-low feminine), feminine (low
masculine-high feminine), androgynous (high masculine-high feminine), and
undifferentiated (low masculine-low feminine). Interestingly though, when the
distribution of subjects based on this new two—dimensional taxonomy are compared with
the distribution that would be expected using the ﬁve gender roles generated from the
BSRI, a largely similar grouping of individuals emerged (see Table 2).

The data prEsented in Table 2 reveal a number of interesting conclusions. First,
although it may be theoretically beneﬁcial to conceptualize gender role identiﬁcation as
two dimensional with masculinity and femininity lying on different continuums
(Constantinople, 1973), it appears that a unidimensional model is a reasonably accurate
substitute in practice. As can be observed in Table 2, approximately 94% of Spence et
al.’s (1975) masculine males and 90% of feminine females fall into Bem’s (1974) near-
masculine/masculine and near-feminine/feminine categories, respectively. Similarly,

approximately 80% of feminine males and 97% of masculine females sorted into the

23

Table 2
Percentage of Subjects in Bem 's (1974) Five Gender Role Categories for each of the
Four Gender Roles Defined by Spence, Helmreich and Stapp (1975)

Spence et al.’s Bern S Categorres

 

 

Categories Feminine Feliiue'arii-ne Androgynous MESZifl-ine Masculine

Males

Masculine (64) 0.0 0.0 6.3 15.6 78.1

Feminine (30) 46.7 33 .3 20.0 0.0 0.0

Androgynous (68) 5.9 1.5 45.6 25.0 22.1

Undifferentiated (72) 8.3 6.9 37.5 16.7 30.6
Females

Masculine (30) 0.0 0.0 3.3 23.3 73.3

Feminine (104) 72.1 18.3 9.6 0.0 0.0

Androgynous (80) 13.8 10.0 47.5 18.7 10.0

Undifferentiated (56) 17.9 19.6 33.9 17.9 10.7

 

Note. Numbers in parentheses indicate the number of subjects within each gender role category
in the Spence et al. study.

Table reproduced from Spence, J. T., Helmreich, R., & Stapp, J. (1975). Ratings of self and peers
on sex role attributes and their relation to self-esteem and conceptions of masculinity and
femininity. Journal of Personality & Social Psychology, 32(1), 29-39.

expected gender role categories proposed by Bem (1974) as well.

Second, it is clear that there is a discrepancy in terms of how to categorize
individuals when they report either high (androgynous) or low (undifferentiated)
identiﬁcation with both gender roles. As Bern (1974) operationalizes the term, an
individual is considered androgynous if the standardized difference score between the
masculinity and femininity subscales is near zero; thus, the majority of people identiﬁed
as undifferentiated or androgynous by Spence et al. (1975) should also be categorized as
androgynous under Bem’s (1974) method. However, as Table 2 indicates, in no case do
more than 50% of undifferentiated or androgynous individuals qualify as androgynous by

Bem’s (1974) standards. It appears then, that either the median split technique utilized by

24

Spence et a1. (1975) is not able to categorize individuals with ambiguous gender roles
reliably or the BSRI’s requirements for establishing androgyny are not sufﬁcient.

Building on the previous point, when subjects were measured using the two-
dimensional model and no dominant gender role emerged (i.e. the undifferentiated and
androgynous groups), the actual gender of the subject was a fairly good predictor of
gender role as conceptualized in the unidimensional case. For example, close to 47% of
males who grouped into Spence et al’s (1975) undifferentiated category were scored as
near-masculine or masculine according to Bem’s (1974) measure. In fact, in only one of
the four cases (androgynous females) did this “actual gender” to “gender role” pattern not
hold consistent. Though these ﬁndings do not necessarily suggest that using gender as a
proxy for gender role is always appropriate when gender role identiﬁcation is ambiguous,
they do suggest that actual gender is a reasonably accurate predictor of one’s “gender-
typica ” characteristics. This holds signiﬁcant meaning when gender role is related to
behavioral outcomes—especially in terms of performance on a traditionally gender-
advantaged task such as a test of mechanical comprehension (Antil & Cunningham, 1982;
see section below on mechanical test performance).

Whether it is conceptualized as one- or two-dimensional, there is little argument
that gender role identiﬁcation has a major inﬂuence on people’s beliefs and attitudes and,
consequently, their behaviors and interests (Bern, 1974; Maccoby & J acklin, 1974;
Spence, Helmreich & Stapp, 1975; Wood & Eagly, 2002). Eccles (1987) summarizes
this point nicely, stating:

“Gender roles mandate different primary activities for men and women. If success in
one’s gender role is a central component of one’s identity, then activities that fulﬁll this
role should have high value and activities that hamper efforts at successfully fulﬁlling
one’s gender role should have lower subjective value.” (p. 152)

25

The reasoning above makes clear two important considerations for the current study.
First, the degree to which gender roles inﬂuence an individual will vary as a function of
the strength and certainty of his/her gender role identiﬁcation (Nash, 1979). Thus, the
ability level of a strong masculine/weak feminine identiﬁed individual and a strong
feminine/weak masculine identiﬁed individual should be signiﬁcantly differentiated,
whereas a less pronounced difference would be expected from individuals with
ambiguous gender role identiﬁcations.

To clarify this point, consider the relationship between mechanical
comprehension and gender role identiﬁcation depicted in Figure 3. As mechanical
comprehension has been shown to be a typically masculine quality (see below, Spence et
al., 1975), it follows that the more strongly masculine-identiﬁed and less strongly
feminine-identiﬁed an individual is the more likely they are to possess higher “levels” of
mechanical comprehension. Alternatively, the more strongly feminine- and less strongly
masculine-identiﬁed one is, the lower their level of the construct. Both of these
statements are represented by the data points labeled “Masculine” and “Feminine” in
Figure 3. Notice also that the level of mechanical comprehension of persons without a
clearly identiﬁed gender role (the “Androgynous” and “Undifferentiated” data points) is
relatively equal. In relation to this study, then, gender role identiﬁcation will be treated
as a two-dimensional, continuous variable in which the most extreme differences will be
found between strongly and wholly gender role identiﬁed subjects.

Second, as Eccles (1987) indicates, the activities a person pursues should be
signiﬁcantly correlated with their most strongly identiﬁed gender role. This should also

necessitate that the interests, knowledge and experiences that an individual accumulates

26

 

(Masculine)

  

(Undifferentiated) . .
,- ._ * ~~' ’ — ' FT .1

Mechanical Comprehension

 

 

 

.
_ - ;" g - T ’ -.- , . L-
j " 0' Masculinity Low (Feminine)
FL fMasculinity l-ligh l i ’ * ” _ * C '*
Low High
Femininity

Figure 3. Example of the hypothesized effects of gender role identiﬁcation on an
individual’s level of mechanical comprehension.

as a result of his/her primary activities should be consistent with their gender role. For
example, one might expect that an individual, regardless of actual gender, who identiﬁes
with the psychological characteristics typical of femininity (i.e. affection, compassion,
sympathy) would pursue activities that are traditionally considered feminine (i.e. cooking,
caring for children), and to thus acquire certain interests, knowledge and experiences that
a masculine-identiﬁed individual would most likely not (Bern, 1974; Wood & Eagly,
2002)

As it pertains to the present study, Spence et al. (1975) presented a number of
descriptive statements to groups of college students and asked them to rate them as either
male-valued, female-valued, or sex speciﬁc (meaning that the “ideal” male or female
would wholly possess/perform the trait/activity, while the other gender would not). Of
the sex speciﬁc items observed by the survey, mechanical aptitude emerged as one of the

characteristics most typical of the ideal male and the masculine gender role. Therefore, it

27

is expected that participating in activities related to mechanical aptitude will be highly

correlated with the masculine gender role, such that:

Hypothesis 3a: Individuals who identify more with a feminine gender role will
have less interest in, less general knowledge of, and fewer experiences with

mechanically-related subject matter.

Hypothesis 3b: Individuals who identify more with a masculine gender role will
have more interest in, greater general knowledge of, and more experiences with

mechanically-related subject matter.

Endorsement of gender stereotypes. Simply stated, a stereotype is a broad
representation of the physical, psychological and behavioral similarities of a large group
of people (Stangor, 2000). Stereotypes about gender, then, can be deﬁned as “socially
shared beliefs about the characteristics or attributes of men and women in general”
(Cleveland, Stockdale & Murphy, 2000, p. 43). For example, Williams and Bennett
(1975) found that 99% of males (n = 50) and 100% of females (n = 50) identiﬁed the
adjective “emotional” as a typical characteristic of women; similarly, 98% of the male
and 98% of the female sample labeled “forceful” a male-typical trait. Once a stereotype
about a particular set of people has been established, its associated characteristics are
often taken as true of any individual afﬁliated with that group (Cleveland et al., 2000,
Stangor, 2000). Thus, based only on the gender stereotype, any female would be
considered emotional and any male would be considered forceful. Such sweeping

generalizations may result in a number of outcomes in the larger societal picture, some

28

more negative (i.e. prejudice, discrimination) than others (Devine, 1989; Lepore &
Brown, 2000).

In focusing more speciﬁcally on gender stereotypes, researchers have captured the
phenomena at two distinct levels of operation—sex—role stereotypes and sex-trait
stereotypes (Williams & Best, 1990). Sex-role stereotypes refer to “beliefs about the
appropriateness of various roles and activities for men and women” (Cleveland et al.,
2000, p. 43), and sex-trait stereotypes refer to “beliefs that [speciﬁc] psychological and
behavioral. characteristics describe the majority of men to a greater or lesser degree than
the majority of women” (p. 44). Based on these more explicit treatments, it is appropriate
at this point to make an important distinction between sex-role stereotypes and the
previously discussed gender role identiﬁcation. Although the two seem very similar, a
major difference exists in how. they are conceptualized, which in turn has signiﬁcant
inﬂuence on their meaningfulness to the current research question.

Sex-role stereotypes are deﬁned as collective judgments of the appropriateness of
the skills, activities, roles, etc. that are associated with a particular gender. On the other
hand, gender role identiﬁcation refers to the degree to which an individual believes the
skills, activities, roles, etc. that he/she possesses are representative of the tasks and
behaviors their sex typically performs (Bern, 1974; Wood & Eagly, 2002). Perhaps at the
risk of oversimplifying the issue, a sex-role stereotype can be perceived as a cultural
directive depicting whether it is appropriate for a male to possess feminine-typical
characteristics and a female to possess masculine-typical characteristics; conversely,
gender role identiﬁcation can be interpreted as a personal endorsement of whether one

actually has masculine or feminine characteristics (irrespective of sex or societal

29

acceptance). This is important in terms of the present research because although sex-role
stereotypes can have self-limiting consequences (Cleveland et al., 2000), such effects
depend greatly on the degree to which an individual actively “buys into” the stereotype
(Eccles, 1987; Nosek, Banaji & Greenwald, 2002) or is aware that a stereotype is
applicable in a given situation (Spencer, Steele & Quinn, 1999). Based on the preceding,
it is reasonable to suspect that gender role identiﬁcation (thought to be a more stable and
enduring individual quality, Constantinople, 1973; Bem, 1974) would be a more valid
and reliable predictor of the individual differences related to sex roles. For the purposes
of this study then, gender stereotypes will only be considered from the perspective of sex-
trait stereotypes.

Well over ﬁfty years of research has attempted to document the traits and
characteristics that are stereotypically associated with men and women, and a large
number of such factors have been identiﬁed that are surprisingly consistent over time and
culture (Williams & Best, 1990). Although these classiﬁcations are interesting and useful
in their own right, the current experiment is more concerned with the individual outcomes
associated with endorsing gender stereotypical beliefs. To this end, Cleveland et al.
(2000) pose an interesting question—might the stereotypical traits of a given sex be
negatively related to the attainment of certain desirable outcomes? Given the wide range
of traits that stereotypically distinguish men from women, researchers have suggested
that there are any number of “male” or “female” behavioral and psychological
characteristics that can explain why various performance, ability, etc. differences exist
(Williams & Best, 1990). For example, achievement motivation has been found to be a

reliable predictor of job performance across a number of studies (Piedmont & Weinstein,

30

1994; Vinchur, Schippmann, Switzer & Roth, 1998; Stewart, 1999; Van den Berg & Feij,
2003). However, Stein and Bailey (1973) demonstrated that men generally exhibit
signiﬁcantly higher levels of achievement motivation than women over the course of the
natural lifespan. Though there may be a number of reasons as to why this difference
exists, Cleveland et a1. (2000) suggest that one possible explanation can be derived by
observing the stereotypical sex traits associated with females in relation to achievement
motivation; across nearly all situations, feminine characteristics (such as non-
assertiveness, avoidance of competition, and dependency) are in direct conﬂict with the
necessary requirements of achievement motivation (e.g. striving for success in all
situations, desire to be “better” than others, etc.). This line of reasoning suggests that the
observed disparity in performance between males and females on a particular task could
be attributed to differences in achievement motivation, which is negatively inﬂuenced by
one’s adherence to traditionally feminine characteristics.

This notion of “compatibility” between feminine/masculine characteristics and
attitudes, behaviors, and cognitions as a driver for the observed gender differences across
various areas of psychological functioning has been labeled role congruity theory by
Eagly and Karau (2002). However, there are two subtle, yet important, components to
the arguments for role congruity theory that are necessary to understand in order to
adequately discuss the present study’s model of gender differences in mechanical
comprehension. First, gender role identiﬁcation is not proposed to directly impact
performance; instead, gender role identiﬁcation affects characteristics of the performance

context (e. g., achievement motivation), which then subsequently affect performance

outcomes (cf. Cleveland et al., 2000). In regards to this study’s model, the intervening

31

variable proposed to be affected is mechanical self-efﬁcacy. Based on the rationales
developed for Hypothesis 2 and Hypotheses 3a/3b, individuals with more masculine
qualities are predicted to possess traits and levels of mechanical interest, knowledge and
experience that predispose them to be more efﬁcacious when it comes to understanding
mechanically-related concepts than more feminine individuals. Further supporting this
reasoning is the ﬁnding reported by Spence et a1. (1975) that mechanical aptitude is
generally perceived as a stereotypically male valued characteristic. Owing to its “male
heritage” then, masculine-identiﬁed individuals should perceive that they are better
equipped (i.e., more role congruous) to deal with mechanically-related activities and thus
report more conﬁdence in their ability to identify and complete mechanical tasks.

The second, less immediately obvious component of this argument is actually an
important caveat to role congruity theory: even in situations in which the characteristics
of a particular gender role are not conducive to a certain favorable outcome (e. g., high
femininity to high achievement motivation), individuals who identify with that gender
role will not always be at a marked disadvantage to their counterparts. For example, a -
signiﬁcant amount of research has been directed at examining the differential outcomes
associated with male versus female leaders (of. Brown, 1979; Eagly, Johannesen-Schmidt,
van Engen & Marloes, 2003). Similar to the point made in the preceding paragraph, the
assumption often examined in the literature is that the qualities which characterize the
feminine gender role are incongruous with those required to be successﬁrl as a leader, and
thus females have a much more difﬁcult time becoming or sustaining leadership positions
(Eagly & Karau, 2002). Although evidence has been found in support of this proposition,

meta-analyses have shown that a number of contextual and individual difference

32

characteristics moderate this relationship such that feminine leaders are not always
passed over or perceived as inferior to masculine leaders (e.g., van Engen & Willernsen,
2004).

What this suggests, then, is that the relationship between gender role
identiﬁcation and performance-related variables can be moderated by a variety of
characteristics at differing levels of analysis. In maintaining the individual-level model
proposed in Figure 2, the moderating variable of interest examined by the current study is
gender stereotype endorsement. The rationale for identifying stereotype endorsement as
a moderator is based on the extensive stereotype threat literature (e. g., Steele & Aronson,
1995; Schmader, 2002; Schmader, Johns & Barquissau, 2004), in which the presence and
salience of a self-important, negative stereotype results in a member of the stereotyped
group underachieving.

In the simplest stereotype threat study, individuals are either told or not told that a
particular test has historically shown signiﬁcant gender, racial, cultural or other group
differences (i.e. a stereotype is made salient). Results ﬁom such studies typically ﬁnd
that in situations where the “threat” or bias is activated, the performance differences
between the stereotyped- group and the neutral-group are much greater than when the
manipulation is not present. Thus the model of stereotype threat assumes that people
enter a situation with differing degrees of interest, knowledge and experience relevant to
the test that might normally predict test performance. But, when the presence of a test
bias is known, this relationship is negatively affected for the predicted group by an

intervening variable. This situation is quite similar to what is hypothesized by the current

33

study, and thus it is theoretically justiﬁable to predict stereotype endorsement as a
moderating variable.

However, unlike stereotype threat research, the present study deals with a
person’s endorsement of gender stereotypes rather than their awareness of the stereotype.
In order to activate a stereotype threat effect, Steele and Aronson (1995) state that an
individual must “face the threat of conﬁrming or being judged by a negative societal
stereotype—a suspicion—about their group’s intellectual ability and competence (p.
797).” In this sense, the stereotype acts only as an arbitrary frame of comparison.
Stereotype threat is not concerned with whether an individual believes the difference
between their subgroup and others is true; instead, it focuses on the processes through
which an individual actively avoids fulﬁlling the stereotype.

This distinction is often presented in the prejudice/discrimination literature as the
difference between stereotype endorsement and stereotype knowledge (Devine, 1989;
Lepore & Brown, 2000). Devine (1989) argues that these two constructs are
“conceptually distinct cognitive structures...[and thus represent] only potentially
overlapping subsets of information” (p. 5). Furthermore, to the extent that both
stereotype knowledge and endorsement are conceptually distinct, they likely have
different implications for the expression and development of related beliefs about the
abilities of oneself and others. In relation to the current study then, the moderating effect
of one’s belief in a stereotype (i.e., endorsement) is predicted to affect performance by
inﬂuencing levels of mechanical self-efﬁcacy, or one’s belief in their mechanical

proﬁciency. This claim simultaneously maintains the well-established SSE-to-

34

performance relationship developed earlier in Hypothesis 2, while still accounting for the
individual inﬂuence of gender stereotype endorsement.

Thus, as the model being tested by the current study follows a basic mediation
model in which mechanical self-efﬁcacy is the link between mechanical interest,
knowledge, and experience and performance on a mechanical comprehension test, it is
hypothesized that endorsement of gender stereotypes will affect performance by
moderating the relationship between one’s gender role identiﬁcation and his/her reported
level of mechanical self-efﬁcacy. As was discussed earlier, gender role identiﬁcation is
closely related to the sex-role stereotype component of gender stereotypes—thus, a
signiﬁcant interactive relationship is expected to emerge between these two constructs
(Figure 4). Though no evidence was found in the literature to support this exact claim, a
signiﬁcant Sex (male or female) by Implicit Stereotype Endorsement (male = math,
female ¢ math) interaction effect on math test performance was reported by Nosek et al.
(2002), indicating that the interaction predicted by the present study is likely.

In summary, it is hypothesized that endorsement of gender stereotypes will
moderate the relationship between gender role identiﬁcation and mechanical self-efﬁcacy,

such that:

Hypothesis 4: A signiﬁcant interaction between gender role identiﬁcation and
gender stereotype endorsement will emerge such that feminine individuals who
more strongly endorse gender stereotypes will report lower levels of mechanical
self-efﬁcacy than those who do not endorse gender stereotypes. The interaction

will not be observed for masculine identiﬁed individuals.

35

 

 

 

 

 

 

 

. .H L _ _ L
> T .T ~ A T T
0 ~
«I ‘ .,
2 . , ,,,~ ‘ #4
it: .
."r ‘ ~ -
8 .,, . ﬂ
a)
To __ _ __ _ __ -
.2
c
I!
§ 1. , W
E W l“ "“" “.ﬂ ’* ____._ “1-.
. —O—Masculrne gender role
_- _ ‘ - .- Feminine gender role LL
Do not endorse Endorse

Gender Stereotype Endorsement

Figure 4. Predicted interaction effect of gender role identiﬁcation and gender stereotype
endorsement on reported levels of mechanical self-efﬁcacy.

Performance on test of mechanical comprehension. The ﬁnal hypotheses made
by the current study concern actually predicting performance on the mechanical
comprehension test. Based on Hypothesis 2, it is predicted that performance will be
directly related to mechanical self-efﬁcacy, a ﬁnding that has been strongly supported by
the literature (Stajkovic & Luthans, 1988; Mathieu et al., 1993; Phillips & Gully, 1997;
Moritz et al., 2000; Lane & Lane, 2001); but as the previous discussions have explicated,
many factors are predicted to inﬂuence mechanical self-efﬁcacy in a number of ways and
combinations, the effects of which should ultimately be seen in overall test performance.
However, none of the studies reviewed included mechanical self-efﬁcacy as a predictor
of mechanical performance, instead opting to look at the singular, causal effects of the

gender-related variables on performance. Thus, in lieu of a more direct correlate of this

36

study’s proposed model, understanding how these various predictors relate directly to
mechanical test performance offers the best means for justifying the performance
predictions hypothesized by the present research.

In one of the earliest studies attempting to capture performance differences based
on gender and gender roles, Milton (1957) compared the scores of 63 males and 66
females with scores on a test of general “problem-solving” skills that had shown
signiﬁcant gender differences previously. Milton succeeded in reproducing the
signiﬁcant overall effect for gender (in which males performed better on the test than
females), but of even greater interest, he also found that sex-role identiﬁcation accounted
for up to 20% of the variance in the male sample’s test scores and 7% of the variance in
the female sample. As further evidence of the importance of sex roles, when scores on
the gender-role identiﬁcation measure were introduced as a covariate in the regression
equation, the previously signiﬁcant sex difference in problem-solving scores dropped to
non-signiﬁcance. Though it is unclear whether the problem-solving skill Milton
measured is speciﬁcally related to mechanical comprehension, these ﬁndings
nevertheless lend signiﬁcant support to the inﬂuence of gender role identiﬁcation on
certain cognitive tasks.

Of a more direct signiﬁcance to mechanical comprehension and the current study,
Nash (1975) conducted an experiment with sixth- and ninth-grade boys and girls to
observe the relationships between gender, gender role preference and spatial ability.
Subjects in the study were asked to complete an open-ended questionnaire regarding
gender role preference that asked them to explain whether they thought it better to be a

male or female and whether they would rather be a male or female. Spatial ability was

37

assessed with scores on the Differential Aptitudes Space Relations Test. Nash found the
consistently reported male advantage in spatial ability that had been noted by other
researchers (Maccoby & Jacklin, 1974); however, an interesting pattern emerged when
gender was crossed with gender role preference (Table 3). Compared by their similar age
groups, sixth- and ninth-grade boys who preferred to be boys substantially outperformed
sixth- and ninth-grade girls who preferred to be girls on the spatial ability test. However,
in both age groups, no signiﬁcant difference in spatial test scores emerged between girls
who indicated they would prefer to be boys and boys who preferred to be boys. This
ﬁnding seems to suggest that the interaction between gender and gender role
identiﬁcation may be a much better indicator than gender alone in predicting performance
on tests of speciﬁc cognitive ability, even in cases where marked gender differences have
been observed. Nash’s study holds importance for the present research question, as
mechanical comprehension has long been known to load on the spatial visualization
factor (Guildford, 1947). Therefore, it is reasonable to expect similar ﬁndings to emerge
when mechanical comprehension is used as the criterion measure.

Antill and Cunningham (1982) present perhaps the most direct—though
Table 3

Mean Spatial Reasoning Performance (SD) on the Diﬁ’erential Aptitudes Space
Relations Test as a Function of Gender and Gender Preference

 

 

6m-Grade 9‘h-Grade 6m-Grade 9m-Grade 6m-Grade 9m-Grade 6m-Grade 9m-Grade
boys who boys who boys who boys who girls who girls who girls who girls who
prefer to prefer to prefer to prefer to prefer to prefer to prefer to prefer to
be boys be boys be girls be girls be girls be girls be boys be boys
(n = 32) (n = 25) (n = 4) (n = 0) (n = 32) (n = 41) (n = 23) (n = 7)
22.96 42.26 6.75 12.91 29.04 21.75 44.19
(15.21) (19.87) (12.20) (13.07) (18.59) (10.59) (17.36)

 

Table reproduced from Nash, S. C. (1975). The relationship among sex-role stereotyping, sex-
role preference, and the sex difference in spatial visualization. Sex Roles, 1 (1), 15-32.

38

incomplete—examination of the effects of gender and gender role identiﬁcation on
mechanical comprehension test performance. Here, the researchers administered ﬁve
different gender-role identiﬁcation measures and the ACER MechaniCal Reasoning Test

to 237 participants (56% female). The sample was then separated into the four gender
role groups speciﬁed by Spence et al. (1975)—masculine, feminine, androgynous, and
undifferentiated—using a median split technique to produce a 2 (gender) x 4 (gender role)
x 5 (measure) matrix. The observed cell means from this experiment are reproduced in

the last four columns of Table 4.

 

 

 

Table 4
Mechanical-Reasoning Means Based on Median Split Method Categories
Category
Tests M + A U + F t M A U F
Males
BSRI 17.1 (68) 15.9 (36) 1.46* 17.0 (46) 17.2 (22) 16.1 (27) 15.2 (9)
PAQ 17.1 (65) 15.9 (39) 1.43* 18.1 (32) 16.1 (33) 16.0 (33) 15.3 (6)
ANDRO 16.9 (79) 15.8 (25) 1.27* 17.1 (46) 16.8 (33) 16.4 (14) 15.0 (11)
CPI 16.8 (78) 16.2 (26) .72 17.2 (49) 16.2 (29) 17.1 (18) 13.9 (8)
Comrey 17.2 (79) 14.9 (25) 2.52*** 17.1 (64) 17.7 (15) 13.8 (8) 15.5 (17)
Females
BSRI 11.9 (51) 11.7 (82) .39 12.5 (19) 11.6 (32) 11.8 (21) 11.7 (61)
PAQ 11.7 (51) 11.9 (82) .35 13.9 (14) 10.8 (37) 12.2 (31) 11.7 (51)
ANDRO 12.4 (51) 11.5 (82) 1.50* 12.0 (20) 12.7 (31) 12.6 (12) 11.3 (70)
CPI 12.6 (49) 11.3 (84) 2.10” 13.8 (12) 12.2 (37) 10.8 (30) 11.6 (54)
Comrey 12.6 (39) 11.5 (94) 1.81" 12.5 (24) 12.7 (15) 11.6 (11) 11.4 (83)

 

* p < .10. ** p < .05 *** p < .01, one-tailed for the directional hypothesis.

Note. Full names of tests are: BSRI = Bern Sex-Role Inventory; PAQ = Personal Attributes
Questionnaire; ANDRO = androgyny scale for the Personality Research F orrn; CPI =
California Psychological Inventory; Comrey = Comrey Personality Scale. Full names of
categories are: M = masculine; A = Androgynous; U = Undifferentiated; F = Ferrrinine.
Numbers in parentheses indicate the number of subjects on which the mean next to it is

based.

The t column refers to the results of the t-test between the M + A and U + F groups in that

row

Table reproduced from Antill, J. K., & Cunningham, J. D. (1982). Sex differences in performance
on ability tests as a function of masculinity, femininity, and androgyny. Journal of Personality &
Social Psychology, 42(4), 718-728.

39

Unfortunately. these authors do not present an explicit test of the gender by
gender role interaction in their results. instead focusing on a broader examination of the
main effects of gender and gender role identiﬁcation. That being said. signiﬁcant effects
were found for both gender and gender role identiﬁcation. As expected. males
outperformed females by a signiﬁcant margin (H1335) = 100.9. p < .001): additionally.
the measures of gender role identiﬁcation also accounted for signiﬁcant variance in
mechanical test performance. even when controlling for gender (F(5.230 = 2.81. p
< .025). An examination of the cell means across gender revealed that masculine
individuals outperformed both feminine and undifferentiated indixiduals in 19 onO cases:
furthermore, in 15 of 20 instances androg}nous individuals also scored higher than these
latter two groups. Also of note, the masculine group outperformed the androgynous
group in 6 of 10 comparisons and the undifferentiated group scored better than the
feminine group in 8 of 10 cases.

Antill and Cunningham (1982) also attempted to examine mechanical test
performance within masculine identiﬁed individuals (ﬁrst three columns of values in
Table 4). The authors began by aggregating the four gender roles (masculine, feminine.
androgynous and feminine) into high-masculine and low-masculine groups for both
males and females separately using a median split technique. The hi gh-masculine group
was constructed by combining the masculine (characterized by Spence et al., 1975, as
high masculine/low feminine identiﬁed) and androgynous (high masculine/high feminine)
participants together, whereas low-masculine merged feminine (low masculine/high
feminine) and undifferentiated (low masculine/low feminine) persons into a single group.

Dividing the sample in this way resulted in the formation of high-masculine/low-

40

masculine groups composed only of males and high-masculine/low-masculine composed
only of females (see Table 4).

t-tests were then performed within gender across the masculine differentiated
groups. As can be seen from their results, the analyses were somewhat inconclusive and
dependent upon the gender role measure implemented. Although the observed mean
scores of the more masculine group were higher than those of the less masculine group in
all but one case, the differences were only signiﬁcant (at the .05 level) in one comparison
for males and in two comparisons for females. Regrettably, although their dataset
allowed for it, a similar analysis was not conducted for high- (feminine + androgynous)
versus low-feminine (undifferentiated + masculine) groups. Had such an analysis been
conducted, the presence of a consistent signiﬁcant difference between high- and low-
feminine males versus hi gh- and low-feminine females would have suggested a possible
gender by gender role interaction. As presented though, the observed differences
appeared to show that higher masculine-identiﬁed individuals performed better on the
mechanical comprehension test than lower masculine-identiﬁed individuals irrespective
of gender, though this conclusion is somewhat suspect given the lack of reliable ﬁndings
across measures of gender role.

Taken together, these ﬁndings reveal important insights into the effect of gender
role identiﬁcation and mechanical comprehension. First, as found in previous studies,
identiﬁcation with the masculine gender role did appear to correlate with the highest
performance on male-advantaged tests of speciﬁc cognitive ability. Second, the Antill
and Cunningham (1982) ﬁndings also suggest that femininity may have a detrimental

effect on mechanical comprehension. In the comparisons where masculinity was a

41

constant (i.e. masculine versus androgynous, undifferentiated versus feminine), the group
that was lowest in femininity was the better performer. Finally, although Antill and
Cunningham did not explicitly test for a gender by gender role interaction, their ﬁndings
do hint that such an interaction may exist. For example, while no analysis was presented
to test for signiﬁcant mean differences across cell means, an examination of their
absolute values (excluding the androgynous and undifferentiated categories) shows that
masculine males were always the top performers, followed by feminine males, masculine
females and ﬁnally feminine females].

While the studies summarized above represent some of the few that have looked
at the effects of gender and gender role identiﬁcation on speciﬁc ability testing, even less
research has considered the role of gender stereotype endorsement on such test
performance. What research is available generally indicates that stereotype endorsement
is not a signiﬁcant main effect predictor of performance. Only in cases where stereotype
endorsement acts as part of an interaction effect does it emerge as a signiﬁcant predictor
of performance. For example, Schmader et a1. (2004) reported that although stereotype
endorsement was not a signiﬁcant predictor of a female’s performance on a mathematics -
test by itself, the stereotype endorsement by stereotype threat interaction was highly
predictive ()6 = -.30, p < .01). Similarly, Nosek et al. (2002) found that the interaction of
implicit stereotypes pertaining to gender and math with gender was a signiﬁcant predictor
of math performance (ﬂ = -.31, p < .01), though, again, stereotype endorsement alone was

not signiﬁcant (,6 = .14, p = ns).

 

' Because males were the high scorers in all instances, the earlier arguments that proposed a biologically-
driven (such as with spatial ability, Harris, 1978; 1981) or socializing (Eccles, 1987) inﬂuence predisposing
men to be better at tasks of mechanical comprehension than women still appear to be strong factors.

42

Based on the preceding, it is predicted that a signiﬁcant three-way interaction
between gender, gender role identiﬁcation and gender stereotype endorsement on

performance will emerge as indicated in Figure 5. Recalling the Spence et al. (1975)

Performance on mechanical comprehension test as a
function of gender and stereotype endorsement with
feminine gender role identiﬁcation

 

 

C a
8 L L a L a
C . 5 ~ - ’D
g . - - ‘ n .
g ‘ q.- L L
0' r ’m ”‘— ’11—“
' -o—Male l

.l :3. ‘Fema'id

 

 

 

Do not endorse Endorse

Stereotype endorsement

Performance on mechanical comprehension test as a
function of gender and stereotype endorsement with
masculine gender role identiﬁcation

 

 

e t -- - ~—
" " O - -’ ."T; _ __ " "m" "
0 L LL. , .LLL .. L . L _ -LL', - Q.-.‘LL__- ,L___._L.
o
c . .- . L ,, __ . LL
f
g . L
a I -"’_T7 "iv ‘1
' —o—-Male
- 0 -Female l
LL- ,LLL LJ-

 

 

 

Do not endorse Endorse

Stereotype endorsement

Figure 5. Predicted three-way interaction effect of gender, gender role identiﬁcation and
gender stereotype endorsement on test performance.

43

ﬁnding that mechanical comprehension is a sex trait stereotypically associated with males
and that gender stereotypes pertain to gender and not gender roles, it should follow that
males will be unaffected by gender stereotype endorsement. However, this should not
hold true in the opposite case; endorsement of gender stereotypes should differentially
affect females such that endorsement results in lower test performance than no
endorsement. In addition, within gender, masculine gender role identiﬁcation should
result in greater test performance than feminine gender role identiﬁcation (Antill &

Cunningham, 1982). In summary then, it is predicted that:

Hypothesis 5: There will be a three-way interaction between gender, gender role
identiﬁcation, and gender stereotype endorsement such that women who are
feminine (or masculine) identiﬁed and endorse gender stereotypes will score
lower than feminine (or masculine) identiﬁed women who do not endorse gender

stereotypes; this pattern will not be evidenced for males (see Figure 5).

As gender stereotype endorsement is predicted to differentially interact with
gender, the remaining hypotheses concern the interaction between gender and gender role
identiﬁcation. To examine this relationship, an attempt will be made to explicitly test the
conclusions implied in the research by Antill and Cunningham (1982), which stated that
masculine males achieved the highest scores on a test of mechanical ability, followed by

feminine males, masculine females and feminine females.

Hypothesis 6a: Males who identify with a masculine gender role will score the

highest on the test of mechanical comprehension.

Hgothesis 6b: Males who identify with a feminine gender role will score lower
on the test of mechanical comprehension than masculine males, but higher than

masculine females and feminine females.

Hypothesis 6c: Females who identify with a masculine gender role will score
higher on the test of mechanical comprehension than feminine females, but lower

than masculine males and feminine males.

Hypothesis 6d: Females who identify with a feminine gender role will score the

lowest on the test of mechanical comprehension.

45

METHOD

Sam le

258 participants from undergraduate introductory psychology courses at a large
Midwestern university enrolled in the study as partial fulﬁllment of course requirements;
Table 5 presents descriptive statistics for the sample. Involvement was voluntary, though
all individuals who participated received class credit. Prior to testing, effort was made to
ensure that the sample was approximately gender balanced in order to ensure that
differences between- and within-gender could be adequately examined. As Table 5
depicts, only 38% (n = 99) of the total sample were male; this slight skew, though, was
expected given the gender distribution of the available subject pool. However, a priori
power analyses predicting a small effect size (d = .2) indicated that 84 to 102 subjects
from each sex (168 to 204 total participants) would be adequate to ﬁnd statistically
signiﬁcant results given this study’s hypotheses thus the unequal cell sizes should not be

problematic.

Measures

Demographics/Background. Standard demographic variables were obtained from
participants in a brief online survey. Items for the demographics/background measure are
presented in Appendix A.

Mechanical interests, knowledge and experiences. Mechanical interests,
knowledge and experiences were assessed using an l8-item survey adapted ﬁom the
Mechanical Experiences Background Questionnaire (MEBQ, Rechenberg, 2000). The

MEBQ is a biodata measure speciﬁcally constructed to assess mechanical trainability. It

46

Table 5

Demographic Information ofStudy Sample (N = 258)

Gender
Male
Female

Age (n = 248)

Ethnicity (n = 255)
American Indian/Alaskan Native
Asian
Black or African American
Native Hawaiian and Other Paciﬁc Islander
White
American Indian/Alaskan Native and White
Asian and White
Black or African American and White

99
159

M= 19.83 (SD = 1.63)

13
32

199

originally consisted of 93 items grouped into 13 dimensions derived ﬁom the job

descriptions and job analyses of mechanically oriented jobs found in the Dictionary of

Occupational Titles. However, as the thesis of the current study is concerned with the

more cognitively-focused construct of mechanical comprehension rather than the more

behaviorally-focused construct of mechanical trainability, many of the MEBQ’S

dimensions are not applicable (such as Following Directions and Coordination). Thus

items were selected from the MEBQ that were judged to most closely conform to the

three dimensions evaluated in the present study (interests, knowledge and experiences).

The wording of items was then adapted such that a 5-point Likert-type scale ranging from

“Strongly Disagree” to “Strongly Agree” could be used to answer each question. Items

for the complete scale can be found in Appendix B.

47

Mechanical self-efficacy. Mechanical self-efﬁcacy was assessed with an 8-item
measure speciﬁcally constructed for the purposes of this research project. To develop the
measures, item content was written to appropriately reﬂect the deﬁnition of mechanical
comprehension presented earlier; namely, a person’s ability to learn about, perceive and
understand physical principles and mechanical elements in everyday situations.
Participants were asked to indicate their conﬁdence level regarding their ability to
perform tasks related to mechanical comprehension on a continuous 7-point scale ranging
from “Not at all confident” to “Completely confident.” The complete mechanical self-
efﬁcacy scale is shown in Appendix C.

Gender role identification. Gender role identiﬁcation was assessed with the Bern
Sex Role Inventory (BSRI, Bern, 1974). The BSRI asks individuals to indicate how
accurately 60 personality characteristics (20 masculine, 20 feminine and 20 neutral itemsz)
describe themselves on a 7-point scale ranging ﬁ'om “Never true of me” to “Always true
of me.” Both a Masculinity and Femininity subscale can then be calculated by combining
respondents’ answers on the respective items. Bern (1974) reports the internal reliability
for scores on the Masculinity items to be a = .86, and a = .80 for items measuring
Femininity. In addition, test-retest reliabilities for both sets of scores were also found to
be quite high (Masculinity r = .90; F emininity r = .90). Items for this measure are

presented in Appendix D.

 

2 Bern (1974) does not make explicit the manner in which the neutral items of the BSRI subscale relate to
items from the masculinity and femininity subscale; given that they are posited to be neutral, the
presumption is that they would not share strong relations. However, an examination of these items in
Appendix D reveals that they likely they do not hold such a neutral valence. The items describe both
“positive” (happy, friendly, helpful) and “negative” (moody, jealous, secretive) aspects of one’s personality,
which, depending upon somebody’s preferred gender role, may load differentially on the
masculine/feminine subscales. Owing to its ambiguous nature then, this subscale is not examined further in
the present study.

48

Gender stereotype endorsement. Gender stereotype endorsement was assessed
with a 10-item stereotype endorsement measure adapted from Blanton, Christie and Dye
(2002) and Levy, Stroessner and Dweck (1998). From Blanton et a1. (2002), participants
are presented with a brief list of common gender stereotypes and asked to rate the degree
to which they personally believe the stereotypes are based on true differences between
males and females. Levy et a1. (1998) proposed ratings be obtained on a 5-point scale
ranging from “Not at all true” to “Absolutely true.” In addition, Levy et al. suggest
providing an example of a true stereotype to reduce a participant’s tendency to answer in
a completely socially desirable manner and report all stereotypes as false. Purposeﬁrl
emphasis was placed on having individuals rate the degree to which “I believe...” a
speciﬁc stereotype is true in order to ensure that the measure was one of stereotype
endorsement and not stereotype knowledge (Devine, 1989; Lepore & Brown, 2000). In a
further attempt to disentangle stereotype knowledge from the measure, the content of the
items was speciﬁcally chosen to represent commonly recognized gender stereotypes, with
the presumption being that the vast majority of respondents should already know about
the purported gender difference and thus stereotype knowledge would be relatively equal
across all participants. Items for the gender stereotype endorsement measure can be
found in Appendix E.

Mechanical comprehension test performance. Mechanical comprehension was
assessed with the most recent version of the Bennett Mechanical Comprehension Test
Form S (BMCT, Bennett, 2006). The BMCT is a timed 30 minute, 68-item, multiple-
choice test published by Harcourt Assessment and is widely considered to be one of the

most frequently administered psychometric tests of mechanical ability (Muchinsky, 2004).

49

The test was speciﬁcally developed to “measure [one’s] ability to perceive and

understand the relationship of physical forces and mechanical elements in practical

situations” (Bennett, 1969, p. 1). The BMCT contains items distributed across 18

categories of commonly encountered mechanical phenomena: Acoustics, Belt Drives,

Center of Gravity, Centrifugal Force, Electricity, Gears, Gravity and Velocity, Heat,

Hydraulics, Inertia, Levers, Optics, Planes and Slopes, Pulley Systems, Resolution of

Forces, Shape and Volume, Structures, and Miscellaneous. Each question consists of a

stem and a corresponding illustration, followed by three response options from which test

takers are to choose the correct answer. Figure 6 provides an example of two questions

taken from an earlier version of the test.

The BMCT manual for Form S (Bennett, 1969) reports published split-half

 

 

 

 

 

 

 

 

 

A B C
If I l l lrl I II J
Alum: rpd Bra (in: g

 

 

LEI Iii

 

 

X

Which man carries more
weight?
(If equal, mark C.)

Y

Which letter shows the seat
where a passenger will get
the smoothest ride?

Source: Bennett, G. K. (1969). Bennett mechanical comprehension test, Form S. New

York: The Psychological Corporation.

Figure 6. Example items from BMCT Form S

50

reliabilities corrected by the Spearman-Brown formula ranging from .81 to .93 across a
variety of samples, with a median value of .86. To obtain an overall score of mechanical
ability, each item is marked correct or incorrect and the total number of correct responses

is simply summed; no subscales are computed from the BMCT.

Procedure

Study participants were recruited from the psychOlogy department subject pool to
take part in the study. In an effort to reduce the amount of time required of subjects, all
surveys except for the BMCT were administered online. Prior to beginning the
experiment, all subjects were required to read and agree with an informed consent form
(Appendix F) that notiﬁed participants of the procedures of the experiment, the
compensation they would receive, and their rights as participants. After providing
consent, all participants completed the online surveys in the same order, ﬁrst responding
to the Demographics questionnaire, followed by the Mechanical Interests, Knowledge
and Experience survey, Mechanical Self-Efﬁcacy survey, the BSRI, and the Gender
Stereotype Endorsement survey. Subjects were then automatically scheduled to come to
a classroom one week later to take a paper-and-pencil version of the BMCT. The one
week break between completion of the online surveys and completion of the mechanical
comprehension test was deemed sufﬁcient enough time to negate any implicit and
inadvertent stereotype threat effects that may have been engaged when ﬁlling out either
the Gender Role Identiﬁcation or Gender Stereotype Endorsement measures (Steele &
Aronson, 1995).

Participants were administered the BMCT in small groups of approximately 10 to

25 individuals. Pencils and scantrons were provided, and the experimenter kept track of

51

the 30 minute time limit on a small handheld egg timer. After all participants had entered

the classroom, test materials were distributed and the following directions provided:

“You will be taking a 30 minute, 68-item test of mechanical ability — please bubble in all
your answers directly on the scantron and leave no marks in the test booklets. When you
are done with the test, return your scantrons, booklets and pencils to the experimenter. If
you ﬁnish the test before the timer goes off, you may turn in your materials and leave.”

During the test, respondents were notiﬁed when there were 15, 10 and 5 minutes
remaining to complete the test. Upon ﬁnishing and handing in their test materials,
participants were offered a debrieﬁng sheet (Appendix G) that outlined the intended

purpose of the study and how the data gathered from the participants would be used.

52

RESULTS
Scale Analyses

Unless otherwise noted, all analyses were conducted in SPSS 13.0. Prior to
hypothesis testing, an extensive examination of the measures used in the study was
undertaken to evaluate the structure and functioning of the scale items. As many of the
scales administered were either created or adapted speciﬁcally for the purposes of the
present research, it was important to determine whether the instruments were operating
correctly and as expected. In order to do so, the response values for all reverse coded
items were ﬁrst changed to ensure that all questions within a scale referenced the
intended construct in the same direction. Following this, scale reliabilities were
calculated to determine if deletion of any of the scale items could improve the overall
alpha level of the measure. Finally, the measures were subjected to exploratory factor
analysis to investigate whether the empirically-derived factor structure corresponded with
the theoretically proposed use of the scales. Based on the interpretation of the factor
structure, any changes regarding the use of the scales in subsequent data analyses were
made accordingly. The results of these efforts are described in the sections below.

It should be noted that a complete scale analysis is not presented for the BSRI and
BMCT. As both of these scales are proprietary instruments with a long history of
normative and speciﬁc instructions regarding their use and interpretation (Bern, 1974;
Bennett, 1969), it was deemed appropriate to use the measures “as is” and thus not
conduct an intensive exploration of the scales. Therefore, only scale reliabilities were

calculated for both, with no subsequent changes made to either measure.

53

Mechanical Interests, Knowledge, and Experiences. Items six, four and four in
the mechanical interests, knowledge and experiences questionnaires, respectively, were
negatively-keyed and thus the response values for these items were changed to coincide
with the remaining survey questions. Reliability estimates were then calculated for each
scale independently. Initially, the alpha levels for the scales were as follows: mechanical
interest, a = .81; mechanical knowledge, or = .73; and mechanical experiences, a = .88.
Examination of the reliabilities at the item level, however, indicated that the removal of
certain items could improve overall scale reliability. Removing items one and six from
the mechanical interests scale resulted in an overall a = .86, removing item four from the
mechanical knowledge scale improved reliability to a = .74, and removing item four
from the mechanical experiences scale achieved an alpha of a = .89.

It should be noted that within all three scales, removal of the reverse coded items
always resulted in a small improvement in reliability, suggesting that respondents did not
answer the negatively-keyed items in the expected direction. Even after recoding the
negatively-keyed items in the correct direction, a subsequent factor analysis using the full
scales revealed an independent factor completely identiﬁed by the reverse coded items.
Based on these analyses, two possible conclusions seem plausible. First, it is possible
that the negatively-keyed questions were poorly constructed, such that they either did not
translate into a direct opposite appraisal of the trait or were tapping a different construct
altogether. However, this seems fairly unlikely given that the three reverse coded items
addressed substantially different content (see Appendix B) and yet still loaded heavily on

a single factor.

54

The second and perhaps more likely explanation is that the questions were not
carefully read and attended to by study participants. Many examples exist in the research
literature demonstrating how careless responding to questions in a survey that contains
both positively- and negatively-keyed items will often result in a separate factor
completely composed of the reverse coded items (cf. Schmitt & Stuits, 1985). As
Schmitt and Stuits (1985) deﬁne it, careless responding occurs when a respondent simply
reads a few of the items in a given scale, infers what the items are asking, and then
responds in like manner to the remaining items without attending to the actual wording of
the questions. When this happens, the reversed wording of the question is not noticed by
the participant and thus the response to that item will not reﬂect a participant’s true
standing on the measured construct. In fact, Schmitt and Stuits found that only 10% of
careless respondents are needed to create a distinguishable, negatively-keyed factor.
Given this evidence, the removal of the negatively-keyed items ﬁom all three scales
seemed statistically justiﬁed in the present context.

Following the removal of the aforementioned questions from each scale, an
exploratory factor analysis utilizing the principal axis factoring model was conducted on
the remaining 14 items from the mechanical interests, knowledge, and experiences scales.
A varimax rotation was then applied to any extracted factors meeting Kaiser’s
normalization criterion. The purpose of this analysis was to determine if treating each
subscale as representative of a single unique construct (i.e., mechanical interests,
mechanical knowledge, and mechanical experiences) was justiﬁed or if instead they
together described one or more different constructs (i.e., mechanical background, etc.).

The results of this procedure are presented in Table 6. Although two factors were

55

Table 6
Unrotated and Rotated (Varimax) Factor Loading Matrix for Items
from the Mechanical Interests, Knowledge and Experiences Subscales

 

 

 

Unrotated Rotated
Items Factor 1 Factor 11 Factor I Factor 11
Mechanical Knowledge #1 .64 .60
Mechanical Knowledge #2 .61 .55
Mechanical Knowledge #3 .43 .39 .41
Mechanical Knowledge #5 .66 .52 .41
Mechanical Experiences #1 .74 .60 .43
Mechanical Experiences #2 .78 .70
Mechanical Experiences #3 .74 .66 .36
Mechanical Experiences #5 .75 .69 .35
Mechanical Experiences #6 .80 .77 .34
Mechanical Interest #2 .65 .32 .62
Mechanical Interest #3 .72 .41 .62
Mechanical Interest #4 .73 .39 .77
Mechanical Interest #5 .71 .33 .33 .71
Mechanical Knowledge #6 .57 .35 .46
Eigenvalue 7.05 1.05 4.04 3.16
% of variance 50.3% 7.5% 28.8% 22.6%

 

Note. Bold values show the highest factor loading. Loadings below .30 are
not presented to ease interpretation.

initially extracted with eigenvalues greater than 1.0, an analysis of the scree plot for the
data indicated that only the ﬁrst factor be retained. However, research (cf. Levonian &
Comrey, 1966; Guertin, Guertin, & Ware, 1981; Rummel, 1970) and theoretical rationale
(cf. Tinsley & Tinsley, 1987) in the broader literature suggest that in cases where the
extraction criteria result in an ambiguous or contradictory statistical conclusion, it is
better to overestimate rather than underestimate the number of factors kept for
interpretation. Therefore, the decision was made to retain both factors for rotation.

Inspection of the loading matrix after rotation revealed that the two factor solution

56

approximated the criteria for simple structure reasonably well. Items on the mechanical
experiences scale seemed to suffer from the highest degree of cross-loading, but in nearly
all cases the higher loadings were sufﬁciently large to justify ascribing the item to a
single factor.

Following rotation, the two factors together explained 51.4% of the variance in
the measures. Factor I, which accounted for 28.8% of the variance following rotation,
was primarily composed of the mechanical knowledge and mechanical experiences items;
Factor 11 accounted for 22.6% of the total variance after rotation and was made up of the
remaining mechanical interest items and a single question ﬁ'om the mechanical
knowledge scale. In order to determine whether any broad themes emerged that could be
used to label and interpret the individual factors, the content of the clustered items was
inspected. Items loading on Factor I appeared to represent an individual’s behavioral,
“outward” engagement with mechanically-related situations and activities, exempliﬁed
by items such as “I typically make repairs around the house when they are needed by
myself rather than ask for help” (Mechanical Experiences #2) and “I am often asked to
show or explain to others how to operate a piece of mechanical equipment (e.g., run a
lawnmower, use a power tool, use a sewing machine, etc.)” (Mechanical Knowledge #1).
In keeping with the previous naming convention, this combined factor was labeled
mechanical experience. The items loading on Factor 11, on the other hand, tended to
focus on a person’s cognitive, “inwar ” engagement with mechanical incidents. Typiﬁed
by questions such as “When faced with an object that isn’t working properly (such as an
appliance or a bicycle), I enjoy trying to ﬁgure out the causes of the malfunction”

(Mechanical Interests #4) and “I try to develop strategies or techniques (e. g. trial and

57

error, working backwards, etc.) for approaching mechanical or technical activities”
(Mechanical Knowledge #6), this factor was most consistent with the previous notion of
mechanical interest.

Table 7 presents the items, descriptive statistics, and reliabilities for the ﬁnal
mechanical interest and mechanical experience factors. As can be seen in the table, the
newly constructed mechanical experience (a = .89) and mechanical interest (a = .85)
scales both demonstrated good overall reliability. On the basis of the preceding analyses,
participants’ responses on the mechanical interest, knowledge and experience subscales
were combined as shown in Table 7, and the average scale scores used for subsequent
hypothesis testing.

Mechanical Self-Efficacy. There were no negatively keyed items included in the
mechanical self-efﬁcacy measure and thus scale reliabilities could be calculated on the
item responses as recorded. Cronbach’s alpha for the 8-item questionnaire was
sufﬁciently high (a = .93), and no improvements to the reliability of the scale could be
achieved by removing any of the items. To determine the underlying structure of the
questionnaire, principal axis factoring was conducted. A single factor solution (2. = 5.43 7)
was supported by the analysis which explained 68.0% of the variance in the mechanical
self-efﬁcacy scale. As a result, no changes were made to the measure and participants’
average scale score was used for hypothesis testing.

Gender role identification. No items in either the Masculinity or Femininity
subscale of the BSRI required reverse coding. The 20-item Masculinity subscale

demonstrated a reliability of a = .86, whereas the 20-item Femininity subscale achieved

an alpha of a = .83. Both of these values are comparable with the reliabilities originally

58

A0080 00.00.00. u m ”00800.00 $800.00. u C 0800 0000080 00808 0 00 00>& 0003000 =< .0002
0:08 9:00:85 8 00000000 000 8800 00.0 5:08:00 ”0000080000 8000.000: 000000000 8 00000000 000 0000: 00.0 80:08:05

 

8000000 80088 00 8080:0000

 

 

m0. Nm.m mm. 82000000 00.0 6000 £08380: 000003 .00000 000 80.0 .000 8:08:08 00 8808.00 00880 00 .9 0
S.— S.m :1. .608 .0800 800:0 w0080 0008.00.30 08:00:00 00003 80300 8800:0000 30: 00.60000 00 M00000 0:: 0
000003800 0:0 .00 80000 0:0 000 000mm
00.0 mad 3.. 00 $00.0 00.0.00 0 A0885 0 00 00002000 00 8 :003 3000000 83003 0.000 8:0 00030 0.0 :03 00000 00:3
.6 000 6000000080 00000 00000000088000:
5: m 0 .m K. 0000: 00 .000. :w 00 805000 8000800 00 80:00:88 800000000 00.0 80:00:02 300 0008 800000. >080 0
.608 .000mb 0080—00800 300
8.: 09m 00. .80—30000000. .000 80000 00880 0800808000 0008 030:0 00800080 00:08? 00 8080800 $00000 02.00 0
mm. 0800000 8008:0002
.6000 .0000 000—00: 0 3008000 .0000 8a 0 388:0
mo: cod mm. .000 0003000 0 w00£ E 0:08: 0003 0:000 00 80:08 8008:0000 .00 08:...» 8000000000 >020 00: 08: 0
mo: cad mm. .0800 000: .00 000 0:0 0000000 8:0 00.80 >090 000000.080 08: _
A9880
8.— mnN Ea m00300 .08000 0000 000.000.5000: 0000: ..w.0v 8003000 800000 00:20 E 0.80 00080 0000000000 08: 0
mo: SN 3.. .08: 00.0 :8 00:0 00:80 080.90 .3 000000 0.8 00:0 00:3 0000: 0:0 000000 008000 00800 E80000 0
no: cod 00. .0000: 0:0 00008 00000 000 0008: 00:0 0 .00 m00500m 83 0 00:3
5. m 0 .m :0. 03000000 $000003 80 00 8:0 000.30 00 :03 w0003 00 8:3 000 000mm 38000 000 0
.6000 000080 8.700000800000000, .0088
am. m _ .N 9.. 88.30 80:00 0 .000 803000 80:00:08 00 8200:0000 00 00880 $008.0 4.02.000 :0000 00: 08: _
.0005 00 02880 a .3 .08 00000.0 .082 .000 0 00.0
1.0 a: on e 320 5020 was .08 80508 080.08. a .8028... a case 0:058 8200020 .088 02 32L 0
.6000 60200.0 80500 a 000 :00. 00300 0 000 003000030—
8: Sum 00. 0 000 .000 0000000000 8080:0000 .00 00000 0 080000 00 .50: 08:8 00 00:08 00 30:0 00 00:8 0000 08 0
0w. 0000000000.: 80:00:00.2
00 as: .0083 as: .

 

80000108800: 00000000000: L000 00:0.20QRM 000.00.000.00: SNNQQ 0.00.00.50Q 0.030000% L000 .3003 60.00.030.082 .008:
0. 03001

59

reported by Bem (1974), suggesting that test takers were responding to the items as
intended. As indicated earlier, no changes were made to either subscale as a result of the
reliability analyses. An average score was thus computed for each subscale and used in
further hypothesis testing.

Gender Stereotype Endorsement. Again, no items in the gender stereotype
endorsement measure were negatively keyed and thus no reverse coding was necessary.
An initial analysis of the lO-item scale revealed acceptable reliability ((1 = .84), and an
item level examination of the measure did not suggest removing any questions from the
instrument. An exploratory factor analysis of the measure’s structure was conducted
using the principal axis factoring model. A varimax rotation was applied to any extracted
factors meeting Kaiser’s normalization criterion. The results of this procedure are
presented in Table 8. Two factors with eigenvalues greater than 1.0 were extracted, a
decision also supported by examination of the scree plot. The rotated factor structure did
not clearly support a simple structure patterning in the data, with items 2, 4, and 6
displaying a fairly high degree of cross-loading.

To determine if a reasonable interpretation of the factor structure could be
achieved, an analysis of the item content was conducted. Items 1, 3, 5, 7 and 10, which
appeared to load rather cleanly on Factor 1, all began by asking respondents to rate the
degree to which they believed “Women are better at than men.” Alternatively,
items 2, 4, 6, 8 and 9, which either demonstrated signiﬁcant cross-loadings or strong
loadings on Factor 2, asked participants to rate the degree to which they believed “Men
are better at than women.” Although the results of the factor analysis do not

support such an unambiguous interpretation of the data, drawing a distinction between

60

Table 8
Unrotated and Rotated (Varimax) Factor Loading Matrix for Items
from the Gender Stereotype Endorsement Scale

 

 

 

Unrotated Rotated
Items Factor I Factor [1 Factor I Factor 11

Stereotype Endorsement #1 .49 .54
Stereotype Endorsement #3 .54 -.38 .66
Stereotype Endorsement #5 .62 -.43 .75
Stereotype Endorsement #7 .67 .56 .37
Stereotype Endorsement #10 .62 .53 .33
Stereotype Endorsement #2 .54 .41 .35
Stereotype Endorsement #4 .62 .44 .44
Stereotype Endorsement #6 .72 .53 .48
Stereotype Endorsement #8 .46 .56 .72
Stereotype Endorsement #9 .64 .53 .81

Eigenvalue 4.10 1 .46 2.57 l .99

% of variance 41.0% 14.6% 25.7% 19.9%

 

Note. Bold values show the highest factor loading. Loadings below .30 are
not presented to ease interpretation.

these two factors in this manner makes sense conceptually. Originally, the measure was
meant to capture the degree to which individuals truly believed there were differences in
ability that could be explained by gender in an absolute sense; in other words, the gender
stereotype endorsement survey was intended to capture an individual’s overall appraisal
of whether ability differences could be ascribed to differences between men and women.
For example, it was believed that a person who responded high to all items on the gender
stereotype endorsement scale was someone who wholly endorsed common stereotypic
ability differences across gender. However, as suggested by this analysis, it appears that
individuals may have been endorsing beliefs about the stereotypic superiority of men and

women separately. Explained differently, the scale seems to be capturing an individual’s

61

endorsement of stereotypes concerning women and their endorsement of stereotypes
concerning men.

Based on the preceding discussion, Factor I (which explained 41.0% of the
variance before rotation and 25.7% after rotation) can be labeled female stereotype
endorsement, and Factor 11 (which explained 14.6% of the variance after rotation and
19.9% after rotation) can be labeled male stereotype endorsement. Table 9 presents the
items, descriptive statistics, and reliabilities for the ﬁnal stereotype endorsement scales.
As shown, an examination of the 5-item subscale for female stereotype endorsement
found an a = .78 while the 5—item male stereotype endorsement subscale achieved an a
= .77, both of which are acceptable levels for the purposes of this research. Additionally,
the factor scores from each extracted factor were not signiﬁcantly correlated (r = .l l, p
> .05), thus reafﬁrming the conclusion that the original gender stereotype endorsement
scale was likely capturing two unique facets. Thus, an average scale score was calculated
for each individual on female stereotype endorsement and male stereotype endorsement.
To examine the effects of these variables in subsequent hypothesis testing, all analyses
that incorporated gender stereotype endorsement were run twice—once using female
stereotype endorsement and once using male stereotype endorsement.

Mechanical Comprehension. No items in the BMCT require reverse coding. The
68-item test had a reliability of a = .83 which is comparable to the median value reported
by Bennett (1969) in the original test manual. As indicated earlier, no changes were
made to the overall scale as a result of the reliability analyses. Each item on the test was
scored correct/incorrect and a ﬁnal total score was computed for the measure to be used

in further hypothesis testing.

62

$3.: $83338 u m ”mg :8 8 B: u 3 280m 889m 8 co 53w 0.8 880%“... =< .802

man—a £53880 8 3:08.: 08 888 com 5:58:9— nmcouﬂohoo 38-80: 382.80 me @3582 08 88: com 32:88:03

 

 

 

:._ SN 3. .8803 .88 £3? 8883. ouch—98 Stop 023 802
o o._ on; av. .5803 .85 35% 8802—88 =80>o 088 28: 802
o: mmd an. .5803 85 mac—m 8883.. 308288 .882. 08: :02
c _ ._ 9mm mm. .5803 82: 25? :88 .8309 028 82
co. cod ow. .5803 .85 mac—m :8QO 5:3 26: 802
2.. 80808085 2502on 032
51— o~.~ em. .58 ~88 tax—Ea 03820 208 08: .8803
.808 55 680 8585008 @888
2 ._ o EN mm. 8 «.8388 85:52 52838 was >208: mini 8 poem 0.8 rod mEo—m Bore—o .5on 28: 5803
m _ ._ m _ .m mm. .808 .85 mad—m .880888 .833 08: 80803
em; and mm. .808 .85 bane 383 539% 08: 8080.3
No; 3 .m mm. .808 .85 ﬁre? 8288:nt 5:3 2:8 5803
ms. 808080c=m 25082m 2889*
am an»: .8583— was—

 

mmNcem. 8853835 eﬁoemxmum. e8: 88 embemhzh eNQEmk .S\«Q.Q 888.:me 3.838% 38 are»: .mmuameeamz 8.88%

a 2.3

3
6

Hypothesis Tests

The means, standard deviations and correlations for all study variables are
presented in Table 10. Additionally, Table 11 gives a comparison of the means and
standard deviations as well as overall effect sizes for each variable across gender.
Signiﬁcant mean differences were found between males and females for all the measured
study variables, with a number of the effect sizes approaching one standard deviation.
The observed performance discrepancy in mechanical comprehension (d = 1.08) was
comparable to that reported in the test manual for the BMCT Form S (d = .8 to 1.2;
Bennett, 1969), which lends strength to the representativeness of the current study’s
sample in relation to the sample from which the normative statistics of the test were
drawn. For the analyses presented below, all continuous predictor variables were
centered prior to entry into the regression equations; thus, all beta weights should be
interpreted as the effect of the independent variable on the dependent variable when all
other variables in the equation are at their respective grand means (Cohen, Cohen, West,

& Aiken, 2003f.

Hypothesis 1. Hypothesis I predicted that individuals with higher levels of
mechanical interest and experience would report higher levels of mechanical self-efﬁcacy.
To test this hypothesis, mechanical self-efﬁcacy was regressed onto mechanical interest
and experience. Table 12 shows that both mechanical interest and mechanical experience
predicted signiﬁcant variance in mechanical self-efﬁcacy (R2 = .45, F (2, 255) = 105.71,

p < .001 ). Examination of the regression coefﬁcients revealed that as the level of interest

(b = .52, t(255) = 4.724, p < .001) and experience (b = .59, t(255) = 6.022, p < .001)

 

3 The dummy coded gender variable was not centered prior to its inclusion in the regression equations.

64

“mob nommgneq8ou 808382 :eSBm
H HOSE $8802: 20% xem 80m mo 283$ 888$ u Ev 3mm $8835 20M. xom 80m .«0 2833 088082 N G6 2mm 682

ﬂu 282 .o n 2883 diets» @088 >883?
2QO 080%2 839w no 88802..
288 080%?— 889m co 38802..

038:9? 80:3 18086 :o 3:82 B .wmm u a no. v Q n .2. .3. v a u .. ”Eon 8 88088 e8 macaw—oboe 882.385

 

 

-- :S. :8. :8... :3..- :2. :3. :3: :N... 3.. mm. 0880 .o
Amwg vor team... .80".- nc. 2.3.. :MN. .136. 3.x 9%; H035 .w
:5 :mm. 8.- :2. 8. no. 8. 8. :8 38888 888% 222 .5
$5 :8. . 8. .2: .2: .2: 3. £8 .88885 ”882% 08:8 .0
$3 8. .2: .3: .2: a. w? .2888 .n
33 :8 :3. :3. 8. 88 3:6 88 .v
33 :3. :3. m: Ea £38833 3288: .m
83 :8. 8. :3” .ooﬁeaxm 83282 .N
E: 2. SM .2293 83282 ._
a w 5 o m e m N _ 8. E 2%.5>

 

83888 85MB 3883.290 VS» waeuuﬁeQ 35885 .28»:

A: 038,—.

65

Table 11
Means, Standard Deviations, and Effect Sizes of Study Variables Across Gender

 

 

 

Females Males
(n = 159) (n = 99)
M SD M SD 1" d

1. Mechanical Interest 2.80 .63 3.41 .64 7486*" .96
2. Mechanical Experience 2.65 .71 3.22 .75 6143*" .78
3. Mechanical Self-efﬁcacy 3.37 1.07 4.33 .97 7280*” .94
4. BSRI (M) 4.96 .62 5.14 .68 2.171“ .26
5. BSRI (F) 5.18 .52 4.66 .56 7675*" -.96
6. Female Stereotype Endorsement 3.02 .84 2.53 .77 4739*" -.61
7. Male Stereotype Endorsement. 2.36 .69 2.64 .88 2.878" .35
8. BMCT 38.35 7.05 46.26 7.60 8502*" 1.08

 

A positive d value indicates that males scored higher than females

*p < .05, ”p < .01, ***p < .001

adf= 256

Note. BSRI (M) = Masculine subscale of Bern Sex Role Inventory; BSRI (F) = Feminine
subscale of Bern Sex Role Inventory; BMCT = Bennett Mechanical Comprehension Test

concerning mechanical-related activities increases, one’s reported level of mechanical

self-efﬁcacy also increases. Thus, this analysis fully supports the predictions made by

Hypothesis 1.

Hypothesis 2. To determine whether the relationship between mechanical interest
and experience on BMCT performance was mediated by mechanical self-efﬁcacy, the

four step regression procedure for testing mediation effects outlined by Baron and Kenny

Table 12
Regression Coeﬂicients for Mechanical Interest and Experience in
Predicting Mechanical Self-Eﬂicacy (Hypothesis 1)

 

 

 

b .8 t” R3
Mechanical Interest .516 .318 4.724***
Mechanical Experience .589 .405 6022*" .45***
nut: p < 001
adf= 255

66

(1986) was used. The results of the analyses are presented in Table 13. In the ﬁrst step
of this procedure, the outcome variable is regressed onto the predictor variable(s) to
demonstrate that a direct effect exists which may be mediated. In the current analyses,
only the relationship between mechanical interest and BMCT scores was signiﬁcant (b =
4.36, t(255) = 4.330, p < .001); the regression coefﬁcient for mechanical experience
failed to achieve signiﬁcance (b = -.44, t(255) = .492, n.s.). Step two requires that the
predictor must also be related to the proposed mediator. As demonstrated in the analysis
for Hypothesis 1, both mechanical interest and mechanical experience were signiﬁcantly
related to mechanical self-efﬁcacy. The third step in the mediation analysis requires that
the mediator be related to the outcome when controlling for the effects of the independent
variable(s); in the present case, the relationship between mechanical self-efﬁcacy and
BMCT scores controlling for mechanical interest and mechanical experience did attain
signiﬁcance (b = 2.68, t(254) = 4.845, p < .001). In the ﬁnal step, full mediation is said

to be achieved if the relationship between the predictor and the outcome drops to zero

Table 13
T est for Mediating Eﬂects of Mechanical Self-Eﬂicacy on Relationship
Between Mechanical Interest and BM C T Scores (Hypothesis 2)

 

‘)

 

 

 

DV Baron & Kenny (1986) Steps b R‘
1. Mechanical Interest 4.361 ***
BMCT
Mechanical Experience -_444 .1 16""
2. Mechanical Interest 516*"
Mechanical Self-Efﬁcacy
Mechanical Experience 589*" .452":
3. Mechanical Self-Efﬁcacy 2683*“
BMCT 4. Mechanical Interest 2.977"
Mechanical Experience -2 .024" . 19 1 Huh

 

*p< .05, **p < .01, ***p< .001

67

when the mediator is introduced into the regression equation; however, in the case that
steps one through three are met but the coefﬁcient does not decrease completely to zero,
partial mediation is indicated. As seen in Table 13, the regression coefﬁcient for
mechanical interest, though attenuated, does not reach zero (b = 2.98, t(255) = 2.956, p
< .01), thus supporting a conclusion of partial mediation. In the case of mechanical
experience, the regression coefﬁcient increased (b = -2.02, t(255) = 2.190, p < .05).

In summary, according to the steps proposed by Baron and Kenny (1986) for
testing mediation, it would appear that only the relationship between mechanical interest
and mechanical comprehension is mediated by mechanical self-efﬁcacy—neither the
initial (Step 1) nor ﬁnal (Step 4) criteria for mediation were met in the case of mechanical
experience. However, an examination of the pattern of regression coefﬁcients and
interrcorrelations among the mediated model’s variables actually suggests that a special
circumstance termed “inconsistent mediation” may actually be occurring (MacKinnon,
Fairchild, & Fritz, 2007). Inconsistent mediation models are models in which at least one
mediated effect has a different sign (+/-) than other mediated or direct effects in the
model. This is evidenced in the present study if one compares the negative regression
coefﬁcient for the path between mechanical experience and BMCT scores in Step 4 (the
direct effect; b = -2.024) to the positive product of the mediating path coefﬁcients in
Steps 2 and 3 (mechanical experience -) mechanical self-efﬁcacy, b = .589, x mechanical
self-efﬁcacy -) BMCT, b = 2.683, = +1.580).

MacKinnon et al. (2007) point out that, although demonstrating the signiﬁcance
of the relationship between the independent (X) and dependent variable (Y) is important

for interpreting the results of the model (i.e., satisfying Step 1 of the Baron & Kenny,

68

1986, procedure), there can be a number of cases in which the X to Y relation is
nonsigniﬁcant though mediation still exists. F urthermore, inconsistent mediation models
' are commonly observed in models such as the present, in which there is more than one
mediated effect being tested (mechanical interest 9 self-efﬁcacy 9 BMCT; mechanical
experience 9 self-efﬁcacy 9 BMCT). In this case, the mediator can actually behave as
a suppressor variable in the presence of two or more moderately correlated independent
variables, thus masking certain expected X to Y relations. As Table 10 shows, the
problem of interrcorrelation among independent variables is a deﬁnite concern in the
presently tested mediation model, in which mechanical experiences and mechanical
interests are correlated at r = .73. Given the potential for inconsistent mediation models,
Kenny (2008) has since revised the original requirement that all four steps of the Baron
and Kenny (1986) procedure be met in order to establish mediation; instead,
demonstrating support for Steps 2 and 3 is considered evidence enough of mediation. In
the present analyses, both mechanical interest and mechanical experience satisﬁed Steps

2 and 3 in the model, and thus Hypothesis 2 was ﬁllly supported.

Hypothesis 3a and 3b. An evaluation of Hypotheses 3a and 3b can be obtained
by examining the correlation matrix presented in Table 10. For Hypothesis 3a,
femininity was predicted to be negatively correlated with mechanical interest and
experience. Evidence was found to support these proposed relationships, with femininity
and mechanical interest obtaining a Pearson r(258) = -.13, r2 = .02, p < .05 while the
correlation between femininity and mechanical experience reached an r(25 8) = -.14, r2
= .02, p < .05. The analyses also yielded ﬁndings in favor of Hypothesis 3b, which stated

that masculinity should be positively correlated with both mechanical interest (r(258)

69

= .34, r2 = .12, p < .01) and mechanical experience (r(258) = .47, r2 = .22, p < .01).

Overall then, both Hypothesis 3a and 3b were fully supported by the present study.

Hypothesis 4. Owing to the multidimensional structure observed in the gender
stereotype endorsement measure employed by this study, an analysis of Hypothesis 4 as
originally described was not pursued. Instead, each proposed relationship was analyzed
twice, once utilizing an individual’s level of female stereotype endorsement and again
with one’s level of male stereotype endorsement.

The ﬁrst portion of Hypothesis 4 proposed a signiﬁcant interaction between
stereotype endorsement and femininity on mechanical self-efﬁcacy; the top half of Table
14 displays the results of this analysis. The regression model including femininity,
female stereotype endorsement and the interaction term accounted for a signiﬁcant
portion of the variance in mechanical self-efﬁcacy (R2 = .05, F(3, 254) = 3.970, p < .01).
However, an examination of the regression coefﬁcients revealed that only the main effect
of femininity was a signiﬁcant predictor of mechanical self-efﬁcacy (b = -.26, t(254) =
2.126, p < .05), indicating that higher levels of femininity predicted lower levels of
mechanical self-efﬁcacy. The model including femininity, male stereotype endorsement
and the interaction term failed to account for a signiﬁcant portion of the variance in
mechanical self-efﬁcacy (R2 = .03, F (3, 254) = 2.483, n.s.).

It was also predicted that reported mechanical self-efﬁcacy would not be affected
by one’s level of stereotype endorsement for masculine identiﬁed individuals; the bottom
half of Table 14 presents the ﬁndings from this analysis. The regression model including
masculinity, female stereotype endorsement and the interaction term accounted for a

signiﬁcant portion of the variance in mechanical self-efﬁcacy (R2 = .14, F (3, 254) =

70

Table 14
Eﬂects of Gender Role Identification and Gender Stereotype
Endorsement on Mechanical SelflEﬂicacy (Hypothesis 4)

 

 

 

 

 

Variable b [9 R2
Femininity (F) -.261* -.l36
Female Stereotype Endorsement (FE) -.136 -.102
F x FE -.201 --.090 .045"
Femininity (F) -.293* -. 153
Male Stereotype Endorsement (ME) .012 .009
F x ME -.112 -.051 .028
Masculinity (M) 595*" .341
Female Stereotype Endorsement (FE) -.202* -.151
M x FE -.141 -.O67 .138**"‘
Masculinity (M) 562*" .323
Male Stereotype Endorsement (ME) -.O35 -.024
MxME .109 .053 .112***

 

*p<.05, **p<.01,p<.001

13.526, p < .001). An examination of the regression coefﬁcients indicated that both the
main effects of masculinity (b = .60, t(254) = 5.831, p < .001) and female stereotype
endorsement (b = -.20, t(254) = 2.581, p < .05) were signiﬁcant predictors of mechanical
self-efﬁcacy, though the interaction effect failed to reach signiﬁcance. The main effects
show that higher levels of masculinity predicted higher levels of mechanical self-efﬁcacy,
whereas greater endorsement of female stereotypes predicted lower levels of mechanical
self-efﬁcacy. The second regression model, which included masculinity, male stereotype
endorsement and the interaction term, also accounted for a signiﬁcant portion of the
variance in reported levels of mechanical self-efﬁcacy (R2 = .11, F(3, 254) = 10.642, p

< .001). Only the coefﬁcient for masculinity (b = .562, t(254) = 5.332, p < .001) was
signiﬁcant, indicating that more masculine individuals reported higher levels of

mechanical self-efﬁcacy overall.

71

Taken together, these ﬁndings only partially support the predictions made by
Hypothesis 4. As predicted, no signiﬁcant interaction between masculinity and
stereotype endorsement (either male or female) emerged; however, the signiﬁcant
interaction proposed to exist between femininity and stereotype endorsement (male or

female) was not observed.

Hypothesis 5. Hypothesis 5 proposed a signiﬁcant three-way interaction between
gender, gender role identiﬁcation, and gender stereotype endorsement on BMCT
performance, with the prediction suggesting a differential effect of stereotype
endorsement. As was the casein Hypothesis 4, Hypothesis 5 was not tested as originally
described because of the two-dimensional structure of the stereotype endorsement scale.
Instead, four regression models were estimated, one for each combination of the 2 gender
role identiﬁcation (feminine, masculine) x 2 gender stereotype endorsement (female
stereotype endorsement, male stereotype endorsement) possibilities.

The ﬁrst regression model estimated included gender, femininity, female
stereotype endorsement, all two-way interactions, and the single three-way interaction in
the equation. Model 1 in Table 15 presents the results of this analysis. The complete
model predicted a signiﬁcant amount of variance in BMCT performance (R2 = .27, F (7,
250) = 13.222, p < .001). An examination of the regession coefficients revealed that
only the main effect for gender was a signiﬁcant predictor of BMCT performance (b =
6.60, t(250) = 6.227, p < .001), indicating that males tended to score higher on the test
than females. It seems noteworthy to point out that both female stereotype endorsement
and the three-way interaction term in this model did attain signiﬁcance at the p < .10

level; however, as neither achieved statistical signiﬁcance according to conventional

72

Table 15

Eﬂects of Gender, Gender Role Identification and Gender Stereotype Endorsement on
BMC T scores (Hwothesis 5)

 

 

 

 

 

Model Variable b a R2

Model 1 Gender ((3)a 6603*“ .392

Femininity (F) -.543 -.039

Female Stereotype Endorsement (FE) -l.3971 -. 144

G x F .931 .043

G x FE -.387 -.023

F x FE -1.028 -.063

G x F x FE 3.6661 .136 .270***
Model 2 Gender (cf 7697*“ .457

Femininity (F) -l. 121 -.080

Male Stereotype Endorsement (ME) -.764 -.073

G x F -.427 -.020

G x ME -.621 -.042

F x ME -.408 -.025

G x F x ME 1.828 .084 245*“
Model 3 Gender ((3)3 6806*" .404

Masculinity (M) .134 .011

Female Stereotype Endorsement (FE) -1.7l4* -. 177

G x M -.275 -.014

G x PE -.703 -.043

M x FE -.336 -.022

G x M x FE -l.108 -.044 .264***
Model 4 Gender (G)a 8.210*** .487

Masculinity (M) -. 141 -.011

Male Stereotype Endorsement (ME) -.947 -.090

G x M .478 .025

G x ME -1.046 -.070

M x ME 1.121 .075

G x M x ME -.389 -.020 242*”

 

*p < .05, ***p < .001, *p < .010
aDummy coded variable, Female = 0, Male = l

73

standards, further examination of this ﬁnding is not presented here (but see the Additional
Analyses section).

The second model regressed BMCT scores on gender, femininity, male stereotype
endorsement, all possible two-way interactions and the single three-way interaction. As
shown in Table 15 (Model 2), this equation also accounted for a signiﬁcant amount of
variance in BMCT scores (R2 = .25, F(7, 250) = 11.608, p < .001). Again, however, only
the regression coefﬁcient for gender (b = 7.70, t(250) = 7.280, p < .001) attained
signiﬁcance, indicating that males outperformed females on the BMCT.

Model 3 in Table 15 presents the results of the regression equation predicting
BMCT performance using gender, masculinity, female stereotype endorsement, all
possible two-way interactions and the single three-way interaction. The model predicted
a signiﬁcant portion of the variance in BMCT scores (R2 = .26, F (7, 250) = 12.826, p
< .001), with the main effects for both gender (b = 6.81, t(250) = 6.947, p < 1.001) and
female stereotype endorsement (b = -1.71, t(250) = 2.510, p < .05) reaching signiﬁcance.
The regression coefficient for gender indicated that males tended to score higher than
females on the BMCT; the coefﬁcient for female stereotype endorsement indicated that
individuals who more strongly endorsed stereotypes about females tended to perform
more poorly on the BMCT.

The ﬁnal regression model (Model 4, Table 15) included gender, masculinity,
male stereotype endorsement, all possible two-way interactions and the single three-way
interaction. The model again predicted signiﬁcant variance in performance on the BMCT
(R2 = .24, F (7, 250) = 11.376, p < .001). Examination of the regression weights revealed

that gender was the only signiﬁcant gender predictor in the equation (b = 8.21, t(250) =

74

8.557, p < .001), showing that males tended to achieve higher scores on the BMCT
compared to females.

In summary, gender was the only consistently signiﬁcant predictor of BMCT
performance across all the tested regression models. However, female stereotype
endorsement did attain signiﬁcance in Model 3 and was marginally signiﬁcant in Model
1, suggesting that individuals who more strongly endorsed stereotypes about females
generally had lower mechanical comprehension scores. Finally, although the three-way
interaction in Model I approached signiﬁcance, the hypothesized interaction between
gender, gender role identiﬁcation and gender stereotype endorsement failed to achieve
signiﬁcance in any of the tested models. Taken together, these results do not support

Hypothesis 5.

Hypothesis 6a, 6b, 6c, and 6d. Hypotheses 6a through 6d sought to explicitly test
the conclusions published by Antill and Cunningham (1982), which stated that masculine
males were the top scorers on tests of mechanical comprehension, followed by feminine
males, masculine females and, lastly, feminine females. Although the median-split
procedure used by Antill and Cunningham to produce their ﬁndings is a common one in
the social sciences, many methodologists discourage the use of the technique because of
the unnecessary restrictions and limitations it places on subsequent statistical analyses.
Speciﬁcally, forcing dichotomization (or any similarly “arbitrary” grouping strategy) is
problematic because it capitalizes on sample speciﬁc data characteristics, requires
perfectly reliable measuring instruments in order to correctly classify individuals at or

around the selected cutoff scores, can generate spurious results, and substantially

75

decrease statistical power (Cohen, 1983; Hunter & Schmidt, 1990; Aiken & West, 1991;
Maxwell & Delaney, 1993).

To circumvent these issues, moderated hierarchical regression is often
recommended as an alternative and preferred statistical procedure (Aiken & West, 1991).
In this technique, the dependent variable is ﬁrst regressed onto the main effects of the two
independent variables. In the next step, the moderator variable (which is simply the
cross-product of the two independent variables) is introduced into the equation; if the
moderator variable predicts signiﬁcant variance above and beyond the two main effects,
an interactive effect is said to be present in the data and a simple slopes analysis can be
performed to determine the pattern of the relationship. Therefore, the same predictions
tested with a median split procedure can be performed without suffering from the
aforementioned deﬁciencies. As such, the present analysis was conducted using
moderated hierarchical regression.

In the present case, because the four gender role categories identiﬁed by Spence et
al. (1975) and analyzed by Antill and Cunningham (1982) were constructed by
combining information concerning one’s level of masculinity and femininity, there were
actually three independent variables to enter into the hierarchical moderated regression
equation—gender, masculinity and femininity. To test Hypotheses 6a-6d, the main
effects for gender, masculinity, and femininity were entered into the regression model
ﬁrst, followed by all possible two-way interactions, and lastly the three-way interaction
term. Support for the predicted gender by gender role interaction is indicated if the three-
way interaction term attains signiﬁcance and further decomposition of the interaction into

simple slopes support the directional predictions made by the hypotheses. Table 18

76

presents the regression coefﬁcients and beta weights of this full regression model. The
results of the analysis show that the three-way interaction term failed to achieve
signiﬁcance (b = -2.98, t(250) = 1.213, p = .23); gender was once again the only
signiﬁcant predictor of BMCT performance (b = 7.78, t(250) = 7.197, p < .001).

In summary, the present study was unable to produce signiﬁcant ﬁndings to
support the results reported by Antill and Cunningham (1982) using hierarchical
moderated regression. Furthermore, even when analyzing the present data with the same
median split procedure originally used by Antil and Cunningham, evidence was not
found to support the original ﬁndings (see Appendix H for a detailed presentation of
these analyses). In both cases, gender was the only variable found to predict one’s scores
on the BMCT, with both gender role identiﬁcation and the accompanying interactions
failing to predict additional variance in mechanical comprehension test performance.

Hence, no evidence was found to support the predictions made by Hypotheses 6a-6d.

Table 16
Eﬁ’ects of Gender and Gender Role Identiﬁcation on
BMC T Performance (Hypotheses 6a — 6d)

 

 

 

Variable b t9 t‘ R2
Gender (cf 7.776 .461 7197*"
Femininity (F) 4.151 -.O83 1.019
Masculinity (M) ”.086 .007 .088
G x F .620 .029 .355
G x M -1552 -.080 .954
M x F -1034 -.052 .607
Gx FxM -2.976 -.110 1.213 491*"
I"!!! p < .001
adf= 250

bDummy coded variable, Female = 0, Male = 1

77

Additional Analyses

Between-Gender Tests. One of the primary goals of the present study was to
speciﬁcally examine and identify potential explanations for the sizable discrepancies in
mechanical ability test performance commonly observed between men and women.
However, large signiﬁcant gender differences were observed not only for scores on the
BMCT, but nearly every variable measured in the present study (see Table 11). This
suggests that the variables captured here might offer additional insight into the
mechanisms that generate the noted performance discrepancies between men and women
if their effects are examined while controlling for participant sex. As such, Hypotheses 1
through 4 (which did not previously include participant sex in the analyses) were run
again, this time partialling out the effects of gender to determine if the observed
relationships still ﬁmctioned similarly.

To reanalyze Hypothesis 1, mechanical self-efficacy was regressed onto
mechanical interest and experience, though this time gender was ﬁrst entered as a control
variable into the regression model (Table 17). In the ﬁrst step, gender explained a
signiﬁcant amount of the variance in mechanical self-efﬁcacy (R2 = .172, F (1, 256) =
53.00, p < .001); the addition of mechanical interest and mechanical experience into the
equation explained an additional 30% of the variance in mechanical self-efﬁcacy (ARI
= .303, AF (2, 254) = 73.08, p < .001). The regression coefﬁcients again showed that,
even when controlling for gender, both mechanical interest (b = .42, t(254) = 3.821 , p
< .001) and mechanical experience (b = .56, t(254) = 5.846, p < .001) were signiﬁcant
positive predictors of one’s reported level of mechanical self-efﬁcacy above and beyond

gender (b = .3 8, t(254) = 3.267, p < .001). Thus the propositions presented in Hypothesis

78

1 were supported: regardless of a person’s sex, greater levels of mechanical interest and
experience predict greater levels of mechanical self-efﬁcacy.

Table 18 presents the results for Hypothesis 2 controlling for gender. Similar to
the ﬁndings reported earlier, mechanical self-efﬁcacy was still found to partially mediate
the relationship between mechanical interest and mechanical experience with
performance on the BMCT. Furthermore, the regression coefﬁcient for mechanical
interests in the ﬁnal step of the mediation analysis dropped even more substantially than
in the previous analysis, so much so that it was no longer a signiﬁcant predictor of
mechanical self-efﬁcacy above and beyond gender and mechanical experiences.
Together, this ﬁnding lends further support to the conclusion posited in Hypothesis 2 that
mechanical self-efﬁcacy is a signiﬁcant mediator of the relationship between one’s
mechanical interest and experience and their mechanical aptitude.

To reevaluate Hypotheses 3a and 3b, partial correlation coefﬁcients were
calculated for mechanical interest and experience with scores on the BSRI masculinity
and femininity subscales excluding the variance accounted for by gender. Contrary to the

previous ﬁnding, when gender variance was removed from the correlation coeﬁicient,

Table 17
Regression Coefficients for Mechanical Interest and Experience in
Predicting Mechanical Self-Eﬂicacy Beyond Gender (Hypothesis 1)

 

 

 

 

Variable b ,6 18? AR"
(Gender) .96] .414": .172 .172":
(Gender) .382 .l65***
Mechanical Interest .424 .261 Me
Mechanical Experience .563 387*“ .474 .303“...
"p < .01, ***p < .001
adf= 254

Note. Gender was used as a control variable in the regression equation

79

Table 18

Mediating Eﬂects of Mechanical Self-Efﬁcacy on Relationship Between

Mechanical Interest and Experience and BMC T Scores Controlling for Gender

 

 

 

 

 

 

 

 

(Hypothesis 2)
DV Step b R2 AR"
BMCT (Gender) 7.910”: .220
. (Gender) 6775*"
Mechanical Interest 2.724“
Mechanical Experience -.902 .248 .028“
Mechanical Self-efﬁcacy (Gender) .961*** .172
. (Gender) 382*“
Mechanical Interest 424*“
Mechanical Experience 563*" .474 .303***
BMCT (Gender) 7910*" .220
. (Gender) 6160*"
Mechanical Self-efﬁcacy 1821*" .272 .052***
. (Gender) 5991*"
Mechanical Self-efﬁcacy 2051*"
Mechanical Interest 1.856
Mechanical Experience -2.057* .290 .017*

 

*p < .05, "p < .01, *** p < .001
Note. BMCT = Bennett Mechanical Comprehension Test
Gender was used as a control variable in the regression equations

Hypothesis 3a was no longer supported; rather than demonstrating a signiﬁcant negative
relationship, near zero partial correlations were observed between femininity and
mechanical interest (rpemqmmsmende, = .06, n.s.) as well as femininity and mechanical
experience (rpemmxpeﬁmwenda = .02, n.s.). However, signiﬁcant positive coefﬁcients

were still observed between masculinity and mechanical interest (rMascqmmsmmc, = .32, p

80

< .001) and experience (rMasceExpeﬁencwende, = .46, p < .001), thus fully supporting
Hypothesis 3b.

The interaction between gender role identiﬁcation and gender stereotype
endorsement (male and female) on reported levels of mechanical self-efficacy above and
beyond the effects of gender was examined for the reevaluation of Hypothesis 4 (Table
19). Although the main effects for masculinity were still strong signiﬁcant predictors of
mechanical self-efﬁcacy even after controlling for gender, no other predictors in the
equation aside from participant sex achieved signiﬁcance. Again, none of the gender role
by gender stereotype endorsement interactions attained signiﬁcance. Thus, the predicted

interaction pr0posed in Hypothesis 4 remained unsupported by the present analyses.

Male-Female Stereotype Endorsement. The separation of the gender stereotype
endorsement measure into subscales that captured beliefs of female and male stereotypes
uniquely opened up the possibility for examining possible interactions between these
distinct constructs. As there was no a priori basis on which to make predictions
concerning their interactive effects, the following examinations were undertaken as an
exploratory investigation of the inﬂuence differential endorsements of male versus
female stereotypes by males and females have on reported levels of mechanical self-
efﬁcacy and mechanical comprehension test performance. Thus, two regression models
were conducted, both of which included gender, female stereotype endorsement and male
stereotype endorsement as predictors in their respective equations.

The full model in which mechanical self-efﬁcacy was regressed onto gender,
female stereotype endorsement and male stereotype endorsement predicted a signiﬁcant

amount of variance in the dependent variable (R2 = .205, F(2, 249) = 21.483, p < .001).

81

Table 19
Eﬂects of Gender Role Identification and Gender Stereotype Endorsement on
Mechanical Self-Eﬂ'icacy (Hypothesis 4)

 

 

 

 

 

 

 

 

 

Variable b ,8 R2 AR)
(Gender) 961*” .414 .172
(Gender) .97 1*" .418
Femininity (F) .046 .024
Female Stereotype Endorsement (FE) -.O36 -.027
F x FE -.210 -.094 .182 .010
(Gender) .961 *" .414 .172
(Gender) 999*" .43 1
Femininity (F) .047 .025
Male Stereotype Endorsement (ME) -.O7O —.048
F x ME -.038 -.017 .174 .002
(Gender) 961*" .414 .172
(Gender) 832*" .358
Masculinity (M) 502*" .288
Female Stereotype Endorsement (FE) -.066 -.049
M x FE -.100 -.047 .253 081*"
(Gender) 961*" .414 .172
(Gender) .901 *** .388
Masculinity (M) .499*** .286
Male Stereotype Endorsement (ME) -.114 -.079
M x ME .020 .010 .254 082*"

 

*p<.05, **p<.01,p<.001

However, the main effect of gender was the only signiﬁcant predictor of mechanical self-
efficacy (b = 1.09, t(249) = 5.902, p < .001), indicating that males tended to report greater
levels of mechanical self-efﬁcacy. Additionally, the main effect of male stereotype
endorsement came very close to reaching statistical signiﬁcance (b = -.34, t(249) = 1.947,
p = .053), indicating that individuals who did not believe that stereotypes about males

were true tended to have higher levels of reported mechanical self-efﬁcacy. Nevertheless,

82

in the present model, the differential endorsement of female or male stereotypes did not
appear to explain differences in one’s reported conﬁdence in their ability to perform
mechanical tasks.

In the second regression equation, a signiﬁcant amount of the observed variance
in BMCT performance was explained by the full model including gender, female
stereotype endorsement, male stereotype endorsement, and their interactions (R2 = .283,
F (1, 250) = 35.088, p < .001). Here, the main effect of gender was again signiﬁcant (b =
8.17, t(250) = 6.527, p < .001) indicating that men tended to score higher than women on
the test. In addition, the main effect of female stereotype endorsement was signiﬁcant as
well (b = -2.17, t(250) = 2.129, p < .05), demonstrating that individuals who more
strongly endorsed stereotypes about women tended to perform worse on the test of
mechanical comprehension. However, both of these main effects were qualiﬁed by a
signiﬁcant three-way interaction between gender, female stereotype endorsement and
male stereotype endorsement (b = -2.92, t(250) = 2.006, p < .05).

To examine the pattern of the signiﬁcant three-way interaction, the two-way
interaction between female and male stereotype endorsement was examined for males
and females separately (Figure 7). For female participants, only female stereotype
endorsement was a signiﬁcant predictor of BMCT performance (b = -2.17, t(155) = 2.160,
p < .05), indicating that women who more strongly endorsed female stereotypes tended to
score lower on the test of mechanical comprehension regardless of their beliefs in male
stereotypes. In the case of male participants, only the two-way interaction term achieved
signiﬁcance (b = -2.61, t(95) = 2.487, p < .05). As can be seen in Figure 7, the pattern of

the data shows that for men who did not strongly endorse male stereotypes, female

83

42

BMCT Score

49.

BMCT Score

411

391

401

47‘
45.

43.

Females

-__.—'.
a ----- *‘--
O"'
.—-""4
”.——P”
r”
Low Mean High

Male Stereotype Endorsement

 

Low Mean High

Male Stereotype Endorsement

- 4. - Low Female
Stereotype
Endorsement

—I— Mean Female
Stereotype
Endorsement

—O-— High Female
Stereotype
Endorsement

- + -Low Female
Stereotype
Endorsement

—I— Mean Female
Stereotype
Endorsement

—0— High Female
Stereotype
Endorsement

Figure 7. Two-way interaction between female stereotype endorsement and male
stereotype endorsement on BMCT performance for females and males

stereotype endorsement made little difference in their overall BMCT performance.

However, male participants’ belief in female stereotypes greatly affected test

performance when they also strongly endorsed male stereotypes; in this case, men who

felt more strongly that female stereotypes are true performed signiﬁcantly worse than

those who did not hold such strong stereotypical beliefs about women.

84

In sum, it appears that female and male stereotype endorsement had little impact
on an individual’s level of mechanical self-efﬁcacy, with males still reporting the greatest
levels of conﬁdence in their mechanical proﬁciency. On the other hand, these beliefs
seemed to have a greater impact on one’s performance on the mechanical ability test.
Males and those who did not strongly endorse stereotypes about women tended to
perform better overall on the test of mechanical comprehension; however, the
performance of males was also dependent upon the degree to which they endorsed

stereotypes about both men and women.

Hypothesis 5 F ollow-up. As mentioned earlier, the model predicted in Hypothesis
5 in which performance on the BMCT was regressed onto gender, femininity and female
stereotype endorsement found that both the main effect for female stereotype
endorsement (b = -1 .3 8, t(250) = 1.888, p = .06) and the three-way interaction term (b =
3.67, t(250) = 1.671, p = .10) approached, though did not obtain, conventional levels of
statistical signiﬁcance (Table 15). While considered poor practice to interpret these
coefﬁcients as “meaningful” ﬁndings based on commonly accepted statistical standards,
given the somewhat exploratory nature of the hypothesis analysis a cursory examination
of their ﬁndings seems warranted. Thus the results of those analyses will be presented
below, with full knowledge that they should not be interpreted as signiﬁcant.

The simple main effect of female stereotype endorsement suggests that
individuals who more strongly endorse stereotypes about females tended to perform
worse on the BMCT. To examine the three-way interaction, the two-way interaction
between femininity and female stereotype endorsement was examined across females and

males (Figure 8). For females, the two-way interaction term (b = -l .03, t(155) = .893,

85

n.s.) indicated that for women who only weakly endorsed female stereotypes, reported
identiﬁcation with the feminine gender role made little difference on BMCT performance;
however, for women who strongly endorsed female stereotypes, low feminine-identiﬁed
women performed slightly better on the BMCT than high-identiﬁed women. For males,
this pattern was quite different. The two way interaction term (b = 2.64, t(155) = 1.374,
n.s.) indicated that when endorsement of female stereotypes was weak, low feminine-
identiﬁed males outperformed high-feminine identiﬁed males on the mechanical
comprehension test. However, and perhaps contrary to what might be expected, for those
who strongly endorsed female stereotypes, high feminine-identiﬁed males achieved

higher BMCT scores than low feminine-identiﬁed males.

86

42-

38

BMCT Score

32»

BMCT Performance

40.

Females

 

Low Mean High

Female Stereotype Endorsement

 

 

Males
. Q
k ‘ ‘ L. n H
k— t x
0 ~ ~ T ‘ ‘ ‘.
Low Mean High

Female Stereotype Endorsement

- -o- - Low Feminine
—I— Mean Feminine
-—o—High Feminine

- + - Low Feminine
—I— Mean Feminine
+High Feminine

Figure 8. Two-way interaction between feminine gender role identiﬁcation and female
stereotype endorsement on BMCT performance for females and males

87

DISCUSSION

The purpose of the present study was to examine the inﬂuence of key individual
difference variables hypothesized to predict the noted discrepancies in mechanical
comprehension across gender. Table 20 presents a formal summary of the hypotheses
pursued in this experiment and the subsequent conclusions supported by the data. The
pattern of results revealed that both mechanical interest and mechanical experiences were
fairly strong predictors of mechanical self-efﬁcacy for both males and females, which
was itself a signiﬁcant predictor of performance on the BMCT. Furthermore, an
individual’s identiﬁcation with male and female gender role characteristics, and
masculinity in particular, was generally indicative of these important mechanical
background variables. Finally, the investigation of gender stereotype endorsement
pursued by this study showed that personal beliefs about the truthfulness of male and
female stereotypes did appear to inﬂuence performance on the mechanical
comprehension test to some degree, a ﬁnding not previously hypothesized or found
within the research literature.

In an attempt to better interpret and ﬂame the present ﬁndings, the following
discussion will broadly focus on an evaluation of the validity of the proposed gender
differences model. More speciﬁcally, this section’s primary focus is on examining which
variables from the model appeared to operate in the manner predicted and which did not,
and subsequently how these ﬁndings contribute to an understanding of speciﬁc cognitive
abilities and the oft observed male-female differences that accompany their measurement.
Attention is also directed towards promoting the applicability of the theoretical/empirical

approach adopted in the current study whereby an individual differences framework is

88

Table 20
Hypothesis Summary

Hypothesis

Result

 

Hyppthesis 1: Individuals with more interest in, greater knowledge
of, and more experiences with mechanically-related subject matter
will report higher ratings of mechanical self-efﬁcacy.

Supported

 

Hypothesis 2: The relationship between mechanical interests,
knowledge and experiences and performance on a test of
mechanical comprehension will be mediated by mechanical self-
efﬁcacy.

Supported

 

Hypothesis 3a: Individuals who identify more with a feminine
gender role will have less interest in, less general knowledge of,
and fewer experiences with mechanically-related subject matter.

Supported“

 

Hypothesis 3b: Individuals who identify more with a masculine
gender role will have more interest in, greater general lmowledge
of, and more experiences with mechanically-related subject matter.

Supported

 

Hypothesis 4: A signiﬁcant interaction between gender role
identiﬁcation and gender stereotype endorsement will emerge such
that feminine individuals who more strongly endorse gender
stereotypes will report lower levels of mechanical self-efﬁcacy than
those who do not endorse gender stereotypes. The interaction will
not be observed for masculine identiﬁed individuals.

Not supported

 

Hypothesis 5: There will be a three-way interaction between
gender, gender role identiﬁcation, and gender stereotype
endorsement such that women who are feminine (or masculine)
identiﬁed and endorse gender stereotypes will score lower than
feminine (or masculine) identiﬁed women who do not endorse
gender stereotypes; this pattern will not be evidenced for males.

Not supported

 

Hypothesis 6a: Males who identify with a masculine gender role
will score the highest on a test of mechanical comprehension.

Not supported

 

Hypothesis 6b: Males who identify with a feminine gender role
will score lower on a test of mechanical comprehension than
masculine males, but higher than masculine females.

Not supported

 

Hypothesis 6c: Females who identify with a masculine gender role
will score higher on a test of mechanical comprehension than
feminine females, but lower than feminine males.

Not supported

 

Hypothesis 6d: Females who identify with a feminine gender role
will score the lowest on a test of mechanical comprehension.

Not supported

 

*Not supported above and beyond the effect of gender

used to model male-female discrepancies in mechanical ability rather than simply

ascribing this disparity to sex. Lastly, the discussion will conclude with a presentation of

the study’s limitations and advice/lessons learned for future related research inquiries.

89

The Gender Difference Model and Specific Ability Testing

As stated earlier, the functional model of male-female differences in mechanical
comprehension proposed by the current study (Figure 2) was designed to examine the
inﬂuence of individual-level, psychological/cognitive variables as predictors of
performance on mechanical ability tests. Additionally, these individual difference
variables were speciﬁcally chosen to capture two distinct characteristics of the participant
thought to be important to the particular cognitive domain at hand. The ﬁrst such
category broadly focused on the gender-related characteristics of the individual (i.e.,
gender role identiﬁcation and gender stereotype endorsement) while the second class of
variables could be classiﬁed as the contextually-related characteristics of the individual
relative to the testing scenario (i.e., mechanical interests, experiences and self-efﬁcacy).
Although the case has already been made for why these particular variables were selected
as important indicators of the gender-related or contextually-related characteristics of an
individual, there is little question thatany number of additional variables could have been
used to represent these psychological aspects of the male and female test taker.
Limitations of content validity aside, however, reviewing the results of this study by
examining the relative impacts of gender-related and contextually-related characteristics
on mechanical ability performance offers a useful and meaningﬁll frame of reference for

interpreting the present ﬁndings and their contribution to the research literature at large.

Gender-related characteristics. The ﬁrst variable considered among the class of
gender-related characteristics was gender role identiﬁcation. As predicted in Hypotheses
3a and 3b, gender role was signiﬁcantly related to mechanical interests and experiences.

However, contrary to what previous research and theory implied (e.g., Bem, 1974; Nash,

90

1975; 1979; Spence et al., 1975; Antill & Cunningham, 1982), no evidence was found to
support any direct or interactive effects involving masculinity or femininity on
mechanical ability (Hypothesis 5, 6a-6d). This non-signiﬁcant ﬁnding was one of the
more surprising results of the study given the nearly identical operational deﬁnition and
measurement approach adopted by the present experiment relative to similar past studies
that did ﬁnd evidence of such a relationship. While it is still somewhat unclear why this
inconsistency was observed, there is an evolving body of research led by contemporary
gender role researchers that proffers one explanation (e.g., Signorella & Jamison, 1986;
Hamilton, 1995; Ritter, 2004). Although their contentions are explicated in greater detail
below, the overall theme of these works posits that the validity of the presumed gender
role-cognitive performance relationship may no longer be appropriate because of
signiﬁcant ﬁlndamental changes in individuals’ understanding of what it means to be
“masculine” and “feminine.” This sociocultural shift, the authors contend, has made the
process of identifying one’s gender role orientation a much more ambiguous task, which
affects the development of a person’s overall self-concept and is ultimately believed to
inﬂuence the attainment of cognitive proﬁciencies and aptitudes.

Originally, the theory behind the gender role-cognitive ability relationship is most
notably attributed to Nash (1979), who posited that individuals, regardless of sex,
perform best on a cognitive task when their level of masculinity and femininity (i.e., their
gender role identiﬁcation) is consistent with the gender stereotyping present in the
cognitive task at hand. Stated more simply, Nash’s (1979) gender role hypothesis claims
that masculine individuals should be more capable of successfully completing

traditionally masculine (as opposed to feminine) cognitive activities and feminine

91

individuals more capable of successfully completing traditionally feminine (as opposed to
masculine) cognitive activities". Implied in this theory is that higher cognitive
performance is achieved because the individual possesses or otherwise “gains” some real
or perceived advantage because of the compatibility between their gender role
characteristics and the task at hand. For example, Massa, Mayer and Bohon (2005)
suggest that the beneﬁt is motivational. Here, the authors propose that individuals exert
more effort and thus perform better on a cognitive task if they believe it taps a knowledge,
skill or ability (“This is a masculine/feminine task”) consistent with their personal gender
role orientation (“1 am masculine/ feminine”) because it reinforces their perception of the
masculinity/ femininity inherent in their self-concepts. Regardless of the mechanism
through which performance is enhanced, however, Nash’s (1979) theory ultimately
suggests that gender role identiﬁcation may be an important component of some
individuals’ cognitive ability performance.

Although numerous examples may be found in the broader research literature that
support Nash’s (1979) general supposition (e. g., Bernard, Boyle, & J ackling, 1990;
Newcombe & Dubas, 1992), an increasingly large number of contradictory ﬁndings are
beginning to accumulate as well (e. g., Hamilton, 1995; Ritter, 2004) in which the link
between gender role and various cognitive ability tasks is either non-existent (as was

found in the present study) or exactly the opposite of what would be expected (i.e.,

 

4 As noted previously (cf. Eccles, 1987), this reasoning is based primarily on social cognitive theories of
development and socialization, in which individuals pursue various experiences and activities speciﬁc to a
given gender role on the basis of the intrinsic and extrinsic reinforcement they receive from them. These
pursuits, in turn, should have a subsequently greater inﬂuence on the speciﬁc proﬁciencies one develops.
This is precisely the framework adopted by the present study’s model.

5 Although the authors did not directly test the motivational aspect of their hypothesis, they did ﬁnd that
whereas masculine women performed signiﬁcantly better on a spatial ability test when they were told it
measured spatial ability, the more feminine identiﬁed women performed better on the exact same test when
instructed the measure actually captured one’s level of empathy. Unfortunately, this pattern of results was
not replicated for men, perhaps suggesting a more complex phenomenon than originally proposed.

92

feminine individuals outperforming masculine individuals on a masculine-oriented task).
A meta-analysis conducted by Signorella and Jamison (1986) offers a particularly
poignant example of this empirical ambiguity. Across the meta-analyzed studies, the
authors noted that higher masculinity and lower femininity tended to be associated with
better performance on mathematical and spatial ability tasks (abilities stereotypically
associated with masculinity), a ﬁnding which would seem to support Nash’s (1979)
theory. Unexpectedly, however, this pattern did not hold consistently across gender; in
fact, their results suggest that adolescent boys who identiﬁed more closely with the
feminine gender role actually tended to exhibit higher mathematical and spatial ability
scores. Furthermore, gender role identiﬁcation did not appear indicative at all of
performance on verbal tasks (a stereotypically feminine-associated ability). Taken
together, although these results do not wholly “disprove” Nash’s (1979) gender role
hypothesis, they do raise some interesting questions concerning the premises on which
the theory is based and the subsequent interpretation of the present research’s ﬁndings.
At its core, Nash’s (1979) gender role hypothesis rests on the assumption that
individuals are capable of successfully performing two key processes: 1) identifying the
“amount” of masculinity and femininity one attributes to their self-concept and thus their
overall gender role orientation, and 2) classifying a particular cognitive domain as
stereotypically masculine or feminine. A breakdown in either of these premises, then,
might account for the non-signiﬁcant relationship observed between gender role and
cognitive performance observed in the present study’s model. It seems unlikely in the
current research, however, that Nash’s (1979) second criterion was violated. As was

pointed out earlier, the association of a given cognitive ability and related tasks with a

93

speciﬁc gender role is largely dictated by cultural norms (Eccles, 1987; Williams & Best,
1990; Cleveland et al., 2000; Wood & Eagly, 2002). Through socialization processes and
experiential learning within a given culture, individuals come to understand and learn
what aptitudes and related activities are typically perceived as masculine and which are
typically perceived as feminine. For example, Huston (1983) and Signorella and Vegaga
(1984) contend that by adolescence, the majority of Americans'have come to understand
that spatial, mechanical and mathematical skills are associated with masculinity, while
verbal skills come to be considered feminine qualities (similar claims were empirically
investigated and supported in earlier works by Spence et al., 1975). As there is nothing
to suggest that the sample used in the current research spanned multiple cultural groups
within which individuals possessed vastly different normative understandings of
masculine versus feminine tasks, it seems safe to assume that the majority of this study’s
participants would consider mechanical ability an aptitude descriptive of the masculine
gender role and could therefore satisfy this component of Nash’s (1979) gender role
hypothesis.

On the other hand, there is some question as to whether individuals are capable of
meeting the ﬁrst premise implied by Nash’s (1979) theory. As Hamilton (1995) points
out, the gender role hypothesis rests on the assumption that individuals are fairly well
gender-typed and would thus exhibit a relatively large difference between their
masculinity and femininity scores on a measure of gender role identiﬁcation. By
extension, this proposition supposes that individuals are clearly in touch with cultural
deﬁnitions of masculine and feminine behavior and that they use such delineations to

evaluate their self-concept and make judgments concerning the appropriateness of their

94

behaviors (Ritter, 2004). However, closer empirical examination of this contention (e.g.,
Twenge, 1997 ; Holt & Ellis, 1998; Duehr & Bono, 2006) has shown that people no
longer appear to identify their personal gender role orientation distinctly enough to
achieve the “boost” to cognitive performance gained from a compatible gender role-
cognitive task match. Likely owing to signiﬁcant sociocultural changes in the 30 plus
years since Bem (1974) and her contemporaries popularized the study of gender role
identiﬁcation, it has become much more socially acceptable for one to develop and
express characteristics indicative of both masculinity and femininity (Twenge, 1997;
Ritter, 2004). In so doing, individuals have become much less likely to deﬁne
themselves as masculine or feminine, instead developing a self-concept that more
substantially blends characteristics of both masculinity and femininity.

For example, a meta-analysis conducted by Twenge (1997) examining changes in
masculinity/ femininity scores on the BSRI and the Personal Attributes Questionnaire
(Spence et al., 1975) between 1973 and 1994 found a signiﬁcant positive relationship
between time and androgyny scores, such that more individuals appeared to be
identifying themselves as high masculine and high feminine as time progressed. This
correlation could partly be explained by the ﬁnding that men demonstrated a minor
positive increase in femininity scores, whereas women exhibited a dramatic increase in
masculinity scores. A longitudinal analysis of gender stereotypes and managerial success
by Duehr and Bono (2006) across a comparable time ﬁame supports a similar conclusion,
reporting that females have become increasingly more associated with typically

masculine managerial qualities.

95

Though Twenge (1997) references research and theories that point to a number of
possible psychological and sociological explanations for this phenomenon, more
important to the present discussion is this trend’s applicability to the results observed by
the current study. As can be seen in Table 10, the mean scores on the BSRI-masculinity
(M) and -femininity (F) subscales for the sample were almost exactly 5.00, indicating
that, on average, participants generally described their self-concept as possessive of
characteristics consistent with both gender roles. Further support for this pattern of
androgyny can be found by examining the gender role scores separately for men and
women (Table 11); although subjects did identify more strongly with their same-sex
gender role (i.e., men more strongly masculine, women more strongly feminine), males
and females still tended to score above the midpoint on both scales. Taken together then,
one explanation for why gender role identiﬁcation did not appear to inﬂuence mechanical
ability in the present study could be that individuals did not perceive themselves as
clearly “masculine enough” to gain any potential psychological advantage in the
stereotypically masculine cognitive domain. Based on this interpretation, these results
seem to support the growing notion that the applicability of the gender role hypothesis in
present day American culture has been minimized.

Despite the above interpretation and the advocacy it has received from some
scholars (e.g., Signorella & Jamison, 1986; Hamilton, 1995; Ritter, 2004), an alternative
view also exists that adopts a much less critical approach of Nash’s (1979) hypothesis. In
this perspective, the validity of the gender role-cognitive ability relationship is not
necessarily discredited; instead, the unreliable relationship between gender role and

cognitive ability is posited as a result of poor operationalization of the primary gender-

96

related constructs involved. As described in detail by Choi and Fuqua (2003), the
traditional method of measuring gender role identiﬁcation (i.e., self-report questionnaires
that ask a respondent how adequately a particular behavioral characteristic describes
himself/herself) does not appear to be an adequate representation of a person’s overall
gender role identiﬁcation. Thus, Nash’s (1979) hypothesis may still be correct, but the
current tools of measurement do not allow us to observe its proposed relationship.

As support of this claim, Choi and Fuqua (2003) summarized the ﬁndings of 23
studies which examined the structure of the BSRI, thus allowing them to describe the
most commonly observed factor analytic pattern to emerge across multiple measurement
instances and samples. Contrary to a simple two-factor (masculinity-femininity)
structure that might be expected and desired, the authors found that anywhere from two
to eleven factors were reported across the various studies. Furthermore, the item loadings
for each factor were not always consistent, varying somewhat based on the sample used
(college vs. non-college populations) and geographic location, among other things.
Nevertheless, the most common factor structure found consisted of four factors: a single
factor composed only of femininity items, two to three factors composed of masculinity
items, and one bipolar factor (commonly labeled a “Sex” factor) on which the two items
“masculine” and “feminine” loaded.

In postulating why this factor structure was so commonly observed, Choi and
Fuqua (2003) suggest that perhaps the subscales of the BSRI are only capturing a very
limited range of the content domain that constitutes “masculinity” and “femininity.” For
example, the most reliable “femininity” factor to emerge consists of only 10 out of the

possible 20 items (affectionate, compassionate, eager to soothe hurt feelings, gentle,

97

loves children, sensitive to the needs of others, sympathetic, tender, understanding and
warm); the remaining items appear to spread out across a number of factors and generate
mostly weak loadings. Echoing the sentiments of other recent researchers as well, they
posit that this lO-item factor could be more accurately described as an
expressiveness/communality factor that occupies only a small portion of the femininity
construct space. In this sense, they argue that the BSRI-F in its current form does not
accurately measure the construct of femininity as a whole. The same contention is
proffered against the BSRI-M, arguing that the two to three most commonly observed
factors reﬂect instrumentality and autonomy rather than the whole of “masculinity.”

As was described previously in the Results section, the present study was not
concerned with examining the validity of the BSRI as a measure of masculinity and
femininity per se and thus the decision to investigate the factor structure of the measure
and subsequently create/modify different subscales was not pursued prior to hypothesis
testing. However, for the purposes of this discussion, a post hoc analysis was conducted
to analyze the factor structure of the BSRI using this study’s sample by subjecting the
twenty items from both the BSRI-M and BSRI-F to a principal factor analysis followed
by a varimax rotation (Table 21). In large part, the results of this investigation revealed a
pattern of item loadings remarkably similar to that summarized by Choi and Fuqua
(2003). Choi and Fuqua (2003) indicate that the single femininity factor most
consistently reported in the literature was composed of 10 items (listed in the preceding
paragraph) from the BSRI-F subscale; these exact same items once again loaded on a
common factor (Factor I) in the present sample, though here an additional item (“Loyal”)

also emerged. Additionally, the bipolar Sex factor (Factor IV) was also replicated in the

98

Table 21
Rotated (Varimax) Factor Loading Matrix for Items from the BSRI

 

 

 

 

 

 

Items Factor I Factor II Factor 1H Factor IV Factor V
Affectionate (F) .75
Compassionate (F) .86
Eager to soothe hurt feelings (F) .65
Gentle (F) .68
Loves children (F) .51
Loyal (F) .42
Sensitive to the needs of others (F) .79
Sympathetic (F) .74
Tender (F) .66
Understanding (F) .62
Warm (F) .70
Ambitious (M) .49
Independent (M) .74
Individualistic (M) .63
Self-reliant (M) .65
Self-sufﬁcient (M) .76
Willing to take a stand (M) .52
Willing to take risks (M) .51
Aggressive (M) .65
Dominant (M) .67
Forceful (M) .71
Feminine (F) .88
Masculine (M) -.84
Acts as a leader (M) .79
Has leadership abilities (M) .61
Eigenvalue 8.32 5.44 2.42 2.03 1.69
% of variance 20.8% 13.6% 6.1% 5.1% 4.2%

 

Note. The letter in parentheses following the item indicates whether it came from the masculine
subscale (M) or the feminine (F) subscale of the BSRI. Factors comprised of items with an
absolute loading less than .4 are not displayed to ease interpretation.

current sample. The three remaining factors consisted of seven, three, and two items,

respectively, from the BSRI-M subscale. These factors did not reproduce a pattern of

99

item loadings similar to that reported in Choi and Fuqua (2003), though Factor 11 could
be described as a combination of the previously observed autonomy and instrumentality
factors. Items loading on Factor III appear to represent the “machismo/alpha male”
mentality commonly associated with highly masculine individuals, whereas Factor V is
clearly related to leadership qualities.

Given the above, what do the ﬁndings from Choi and Fuqua (2003) and those
presented in Table 21 mean in relation to this study’s results and the future of gender role
research? To begin, one obvious implication concerns the use of the BSRI as an accurate
measure of gender role identiﬁcation. In a lengthy discussion of the ten greatest
problems facing the measurement of masculinity and femininity, Beere (1990) points out
that issues of content validity and multidimensionality in currently developed measures
of gender role identiﬁcation continue to be two of the most proliﬁc challenges facing
gender role researchers. The BSRI is clearly no exception to this criticism; to assume
that 20 behavioral characteristics are sufﬁcient to describe a construct as complex as
masculinity or femininity, especially considering their dependence on oﬁen changing and
ambiguous cultural deﬁnitions, seems a stretch. What’s more, even with the limited
number of items currently in use, multiple factors still managed to emerge from the
purportedly unidimensional subscales (Bern, 1974). Thus as it currently stands, the
present research supports the notion that the BSRI does not appear to be as valid a
measure of masculinity and femininity as it was originally intended.

On the other hand, though, these ﬁndings could also be taken to suggest that
masculinity and femininity should be treated as higher order latent variables that are not

directly measurable but which could be estimated from more proximal latent subfactors

100

using structural equation modeling techniques. To the best of this author’s knowledge,
such a consideration has not been explicitly examined or considered in the research
literature; nevertheless, given the results presented above, it appears a plausible
alternative. In this sense, one could argue that the BSRI is useful for capturing a select
few lower order factors (i.e., expressiveness, instrumentality, etc.) that describe a portion
of the unique construct spaces deﬁned by femininity and masculinity. If this were the
case, one could then interpret the non-signiﬁcant relationships observed in Hypothesis 5
and 6a-6d by stating that the speciﬁc facets of masculinity and femininity measured by
the current study did not appear to inﬂuence mechanical ability.

The above conclusion does not necessarily negate the predictions of the gender
role hypothesis proposed by Nash (1979), but instead qualiﬁes it by stating that only
certain aspects of one’s gender role orientation provide a “boost” when performing a
gender stereotyped cognitive task. To identify and validate the notion of multiple lower
order masculinity and femininity factors would require extensive efforts directed at item
development in order to accurately sample the content domain of each gender role
construct, followed by repeated factor analyses across multiple samples. However, if
such facets could be reliably identiﬁed, the boon to future research could be substantial—
much in the same way the creation of the “Big 5” reenergized personality research, the
same could be achieved in the ﬁeld of gender role research. Through continued research
on and development of measures for lower order masculinity and femininity facets, a new
and better understanding of gender role orientation might be achieved that would be
much better suited for examining speciﬁc gender hypotheses that are currently not

adequately explained.

101

An additional brief note of interest concerns the bipolar Sex factor identiﬁed by
Choi and Fuqua (2003) and reproduced with this study’s sample. Although
Constantinople (1973) popularized the notion that masculinity is not simply the opposite
of femininity (and vice versa) and thus gender role identiﬁcation should not be placed on
a continuous unidimensional spectrum, the observed Sex factor suggests that individuals
do in fact consider these two descriptors in rather mutually exclusive terms—more often
than not, respondents believed either masculinity or femininity was the better descriptor
of their self-concept, but not both. However, as previous research has shown (e.g.,
Twenge, 1997) and the current study replicated, individuals are quite likely to endorse
characteristics from both the BSRI-M and BSRI-F and therefore be classiﬁed as high
masculine/high feminine.

How, then, can these seemingly contradictory ﬁndings be reconciled? While a
deﬁnitive answer can not be derived on the basis of the present results, the argument
could be made that the Sex factor is merely an artifact of language and represents
respondents’ conscious effort to avoid the cognitive dissonance that could occur from
positively endorsing perceived antonyms in a measure. For example, although not
presented in Table 21 because of their small factor loadings, the masculine items
“analytical” and “athletic” both grouped on a single bipolar factor as well, suggesting that
respondents tended to think of these as opposing descriptors of their self-concept. Given
popularly portrayed stereotypes and images of the “dumb jock” and the “frail
bookworm,” it does not require much imagination to guess why individuals might
consider these two terms incompatible and thus endorse only one or the other as part of

their own self-concept.

102

In similar fashion, perhaps individuals felt uncomfortable stating they were
simultaneously masculine and feminine knowing that the two terms are often associated
with seemingly opposing personal characteristics. These aspects of social desirability are
deﬁnite issues of concern when using self-report, personality-type questionnaires—
though little is known about how social desirability affects participants’ responses on the
BSRI. Although the BSRI contains a social desirability subscale, it is unclear how this
factor is used (or was intended to be used), and a number of the items are likely not as
“neutra ” as they were intended (see Appendix D). However, both of these conjectures
are speculation at best; a more explicit examination of the Sex factor and social
desirability in the BSRI is needed in future research to better understand how these issues
affect the construct validity of the measure.

The second and ﬁnal gender-related characteristic examined in this study’s model
was gender stereotype endorsement. Stereotype endorsement had previously been
identiﬁed as a signiﬁcant moderator of relationships between various individual
difference variables and performance-related outcomes in other works (e. g., Levy et al.,
1998; Blanton et al., 2002; Schmader et. al., 2004), and thus was included as a moderator
in the present analyses as well. However, the endorsement of male or female stereotypes
did not inﬂuence mechanical self-efﬁcacy (Hypothesis 4) or performance on the
mechanical comprehension test (Hypothesis 5) in the manner predicted, making an
interpretation of this variable as a useful predictor of gender differences difﬁcult. As was
the case with gender role identiﬁcation, this issue of non-signiﬁcance again begs the
question: Why were similar signiﬁcant relationships not observed in the present

experiment though strongly suggested by previous research?

103

One potential explanation centers on the operationalization of gender stereotype
endorsement in general, and the measurement strategy employed in the current study in
particular. Although this study attempted to use a nearly identical measure of stereotype
endorsement as that implemented by previous researchers, the issue of measurement
contamination may have been a legitimate concern here and one that could have
inﬂuenced the outcomes observed in this study. Similar to past research, the present
experiment deﬁned stereotype endorsement as the degree to which an individual believes
a particular characteristic about males or females is true. To operationalize this deﬁnition,
items in the measure were constructed so as to read in the following manner: “Men have
better skills than women” (or “Women have better skills than men”).
In some respects, indicating whether this item truthfully reﬂects a characteristic of men
(or women, as the case may be) does appear to satisfy the intended nature of the measure;
however, this structure also adds an unnecessary comparative component by forcing a
frame of reference on respondents. By deﬁnition, stereotypes are merely characteristics
mentally associated with a social category (Stangor & Lange, 1994); they suggest nothing
about whether the same characteristics can or can not apply to a greater/lesser degree for
other social categories. In this sense then, stereotype endorsement may imply a relative
judgment, but it does not explicitly require one. For example, the same question ﬁ'om
above might have been re-written as: “On average, men are very good at things that
require skills.” Responding to the truthfulness of this item still provides an
indication of one’s belief in the validity of the particular stereotype at hand, but does so

without imposing an explicit comparison between males and females.

104

The reason this comparative framework is potentially worrisome is because it
introduces the very real possibility of a confounding effect based on individuals’ desire to
promote positive views of their social group, which in this case refers to their sex. Tajfel
and Tumer’s (1986) work in social identity theory posits that feeling as though one
“belongs” to a particular social network or group of like-minded people provides one of
the primary sources of reinforcement for individuals. Similar to in-group/out-group
theories of social comparison, people attribute the positive characteristics of their social
groups to their own self-concepts and thus beneﬁt from identifying with a group that
possesses desirable qualities. In turn, people should be motivated to advocate that the
groups they belong to are generally “superior” (or, at the very least, are no worse off)
than other groups as a means of improving, maintaining or protecting their own positive
sense of self. In the present study, the only groups made explicitly salient to respondents
were males and females. Thus, asking women to directly endorse a stereotype that males
are better at something than females is, according to social identity theory, akin to asking
them to ﬁeely admit their social group is of a lower status (the opposite would be true for
males as well).

Based on this logic, one would expect men and women to most strongly endorse
items that reﬂect positively on their own sex. As all stereotype endorsement items in the
present study were worded in the afﬁrmative (i.e., Men/women better than women/men
at ), this response pattern would be reﬂected in the data as females reporting
higher ratings on the female stereotype endorsement scale than males (thus more strongly
endorsing pro-female stereotypes than males) and males reporting higher ratings on the

male stereotype endorsement scale than females (thus more strongly endorsing pro-male

105

stereotypes than females). As can be seen in Table 11, this is precisely the response
pattern observed in the present study. On the female stereotype endorsement scale,
women tended to score d = .61 standard deviations higher than men, while men tended to
score d = .35 standard deviations higher than women on the male stereotype endorsement
scale, both of which indicated signiﬁcant mean differences. Taken together, these results
appear to be strong evidence that respondents in the present sample were answering in a
self/group-reinforcing manner predicted by social identity theory (Tajfel & Turner, 1986)
rather than objectively evaluating the validity of a particular stereotype.

How, then, might this ﬁnding explain the non-signiﬁcant effects of stereotype
endorsement reported here compared to studies in which the variable was a signiﬁcant
predictor of similar cognitive performance outcomes? For one, the majority of studies
reviewed that employed similar measures of explicit gender stereotype endorsement were
attempting to elicit social comparison or stereotype threat in their samples (Levy et al.,
1998; Blanton et al., 2002; Schmader et al., 2004). Thus, any self/group—reinforcing
response pattern that the measure engendered was to be expected and actually desired, as
it would then be seen as evidence that the experimental manipulation (e. g., making
gender salient to participants, drawing attention to one’s group versus personal identity,
etc.) was successful. Therefore, in these studies, the pro-social group response pattern
would not be perceived as confounding.

However, in the present research, social comparison was not directly manipulated
nor was it a desired characteristic of the experimental context, and thus the observed
gender differences in the stereotype endorsement measures could not be attributed to

controlled variation in the study. As a result, the non-random, non-normal variance

106

observed in both endorsement measures likely violates the assumption of residual
independence; in other words, there is strong reason to believe that being male or female
was signiﬁcantly related to the responses individuals made to both the male and female
stereotype endorsement measures (Cohen, Cohen, West & Aiken, 2003). Although
nonindependence of the residuals does not affect estimates of regression coefﬁcients, it
does affect calculation of the standard errors, which in turn increases the potential for
Type 11 errors. In sum, the fact that gender was indicative of responses to the stereotype
endorsement measures likely lessened the probability that any relationships involving the
construct would reach statistical signiﬁcance (for example, the near-signiﬁcant
interaction observed between gender, femininity and female stereotype endorsement in
Hypothesis 5, see Figure 8).

Despite the issues associated with poor operationalization and the resultant
confounding effect caused by the self—/group-reinforcing response pattern, the present
experiment did identify an interesting interactive effect between male and female
stereotype endorsement across gender on performance on the mechanical comprehension
test (see Male-Female Stereotype Endorsement in the Additional Analyses section). To
summarize brieﬂy, a main effect of female stereotype endorsement was found for females
such that higher levels of endorsement predicted poorer test performance. However, for
males, a signiﬁcant two-way interaction emerged between male and female stereotype
endorsement such that men who strongly endorsed both male and female stereotypes
performed signiﬁcantly worse on the test compared to men who more strongly endorsed

male stereotypes as opposed to female stereotypes.

107

Two tentative insights can be drawn on the basis of these ﬁndings. The ﬁrst
concerns the inverse relationship observed between female stereotype endorsement and
BMCT test performance. As the analysis of Hypothesis 5 demonstrated, the negative
main effect of female stereotype endorsement on mechanical ability either approached or
attained statistical signiﬁcance in both regression equations for which it was estimated
(see Table 15), lending a moderate degree of stability to the result reported above. But
why would one’s belief in the validity of stereotypes about women inﬂuence performance
on a test of mechanical ability? In the introductory sections in which the justiﬁcation for
the proposed gender differences model was presented, the argument for including gender
stereotype endorsement as a meaningful variable was based on the belief that it produced
an effect similar to stereotype threat (e. g., Steele & Aronson, 1995; Schmader, 2002;
Schmader et al., 2004). Given the preceding discussion on social identity theory (Taj fel
& Turner, 1986), the endorsement of female stereotypes might also be perceived as an
indirect measure of one’s identiﬁcation with the female sex. Together, these two
interpretations could be taken as evidence to suggest that individuals who strongly
endorsed female stereotypes were more fully aware of their “femaleness,” and thereby
experience a stereotype threat-like effect in which this female quality was not seen as
conducive to performing well on a cognitive ability test typically associated with male
superiority. While this ﬁnding has been somewhat elusive in previous research, the
notion that gender identiﬁcation can play a signiﬁcant role in the saliency and potency of
stereotype threat effects on test performance has been empirically demonstrated

(Schmader, 2002).

108

Without question, this argument rests on the debatable assumption that the
endorsement of male/ female stereotypes is signiﬁcantly correlated with one’s level of
gender identiﬁcation with the male/female sex. It would certainly seem plausible to
imagine a scenario in which a highly female identiﬁed individual would not endorse
positive stereotypes regarding the female gender. Nevertheless, the implications from
Taj fel and Turner’s (1986) social identity theory would seem to favor the prediction that
a relatively strong relationship exists between gender stereotype endorsement and gender
identiﬁcation, especially if the stereotypes being endorsed are largely positive in nature
(as in the present study). Again, if individuals seek reinforcement ﬁom their group
identiﬁcation as the theory suggests, they should be more likely to endorse positive
stereotypes about their group to reinforce their personal belief structures. Unfortunately
the present research is unable to test this supposition as gender identiﬁcation was not
included among the study’s measured variables; future research would be needed to
clarify whether this relationship can be empirically reproduced.

Although the above claim is somewhat speculative, what makes it a potentially
intriguing interpretation is that whereas stereotype threat is a function of the experimental
context and caused by a purposeful manipulation from the researcher, the effects of
stereotype endorsement come directly ﬁ'om the implicit cognitions or attitudes of an
individual (N osek et al., 2002). By this deﬁnition, then, one might alternatively perceive
stereotype endorsement as a “self-constructed” stereotype threat effect. In other words,
because stereotype endorsement is proposed to be a direct outcome of a self-evaluative
process in which respondents tap into past experiences, cultural norms, personal beliefs,

etc. in order to assess the validity of a given stereotype (Schmader et al., 2004), the

109

individual could be viewed as the primary cause for creating the threatening situation.
From a practical standpoint, such a possibility has very real implications for the strategies
one might pursue to reduce gender differences in test performance and the subsequent
development of more “gender-ﬁiendly” tests. Although stereotype threat research would
suggest that the removal of any material that could potentially arouse awareness of
gender stereotypes or one’s gender identity from a testing situation would
prevent/alleviate large sex differences in test performance, the above interpretation
implies that people carry these pieces of information around with them all the time and
could therefore be aware of them regardless of the contextual information present in the
test environment. Thus, a better way to minimize gender differences in such situations
might be teaching individuals test taking strategies that reduce the saliency/importance of
one’s gender identity in relation to test performance and thus prevent invoking possibly
debilitating thought patterns. However, as the present research does not speciﬁcally test
nor do its results speciﬁcally support such a notion, future research would be needed to
examine the validity of this claim.

The ﬁnal point of interest in relation to the three-way interaction found between
gender, female stereotype endorsement and male stereotype endorsement is closely
related to the discussion above, though it concerns the signiﬁcant two-way interaction
observed for male participants between female and male stereotype endorsement. Again,
the shape of this interaction was such that although female stereotype endorsement had
no effect on test performance for low male endorsing men, high male endorsing/high
female endorsing individuals did signiﬁcantly worse on the mechanical ability test than

high male endorsing/low female endorsing individuals. An interpretation of this ﬁnding

110

does not follow clearly from the previous conceptualization of stereotype endorsement as
an indirect indicator of gender identiﬁcation (Tajfel & Turner, 1986) and a self-imposed
stereotype threat effect. Males who strongly endorsed female stereotypes (indicating an
identiﬁcation with the female sex) only appeared to do poorly on the mechanical ability
test when their level of male stereotype endorsement was high as well (indicating some
identiﬁcation with the male sex as well); when female endorsement was high, but male
stereotype endorsement was low, performance on the test was not attenuated.

Unfortunately, the broader research literature on stereotypes and social identity
offers little help in suggesting why these results were observed in the present study.
Perhaps one explanation, however, could be surmised by considering the possible
situational demands associated with male performance on a test of mechanical ability.
One intriguing line of stereotype threat research suggests that males perform better on a
stereotypically male-advantaged/female-disadvantaged cognitive ability test (such as the
BMCT) through the process of stereotype lift (Walton & Cohen, 2003). In brief,
stereotype lift proposes that a performance boost may be experienced by members of a
particular social in-group when they are made aware of a negative stereotype about the
abilities of an opposing out-group which can subsequently be used as a base of
comparison. Similar to the rationale behind a self-constructed stereotype threat effect
then, a similar self-constructed stereotype lift effect might exist in which the information
required to identify one’s relative membership status and their comparative standing
versus the out-group is already possessed by an individual.

If this were the case, stereotype lift would be demonstrated in the present sample

if high male endorsing/low female endorsing men (i.e., “male afﬁliated” men) performed

111

better than a low male endorsing/low female endorsing or low male endorsing/high
female endorsing male (i.e., not “male afﬁliated” men). Although this precise ﬁnding
was not statistically signiﬁcant in the present dataset, there did appear to be at least some
evidence to suggest such a phenomenon could be taking place. A close examination of
Figure 7 does show a slightly positive slope for low female stereotype endorsing men as
they more strongly endorse male stereotypes. Again, the results from this study do not
provide cOnclusive enough evidence to support this conjecture, but perhaps with a larger
sample of males (n = 99 men in the current sample) a more pronounced and statistically
signiﬁcant difference would emerge.

How, though, might the rationale behind stereotype lift explain the poor test
performance of high male endorsing/high female endorsing men in relation to other males
in the sample? The argument is once more speculative in nature, but social identity
theory offers one possible solution to this quandary (Tajfel & Turner, 1986). Although
this portion of the male sample tended to report in a self-/group-reinforcing pattern that
indicated their “maleness,” they also were also equally aware of their “femaleness.” This
being the case, it could be argued that such high male endorsing/high female endorsing
men would experience some manner of cognitive dissonance if they were to beneﬁt ﬁ'om
their “rnaleness” by denigrating their “femaleness.” Their identiﬁcation with the female
sex and the associated dissonance could have thus resulted in the observed pattern of
diminished performance. Although circuitous support at best, it is interesting to note that
high male endorsing/high female endorsing men achieved performance levels on the
BMCT (M = 40.0) comparable to those reached by the top-performing females (M =

41.5). In this sense, one could argue that high male endorsing/high female endorsing

112

males still performed relatively well in relation to the majority of females (thus satisfying
the quality of their “rnaleness” that implies they should do well on a test of mechanical
ability), but without the advantage of stereotype lift that would have been gained by
debasing the strongly felt “female” portion of their self-concept (thus maintaining a
positive perception of their “femaleness”).

In summary, the gender-related characteristics of the proposed gender differences
model did not generally inﬂuence mechanical ability test performance or mechanical self-
efﬁcacy in the manner in which they were hypothesized (Hypothesis 4, 5, and 6a-6d).
Upon closer examination though, a number of intriguing alternative explanations and
implications were observed that suggest potentially new and as yet unexplored avenues of
research. However, a cautionary word of warning is warranted at this point against
placing too much credence in the theoretical interpretations proffered above without
further and more deliberate empirical investigation. Nevertheless, continued exploration
of gender-related variables such as gender role identiﬁcation and gender stereotype
endorsement would undoubtedly improve researchers’ and practitioners’ understanding
of what it “means” to be men or women in relation to describing male-female differences,
and ultimately allow for a better appreciation of what and how other gender associated

inﬂuences affect cognitive ability test performance.

Contextually-related characteristics. With only the one exception noted above
(Hypothesis 4), the remaining hypotheses examining the contextually-related
characteristics of the gender differences model (Hypotheses l, 2, 3a/3b) were well
supported. This suggests that mechanical interests, experiences and self-efﬁcacy play

integral roles not only as predictors of mechanical ability, but also as meaningful sources

113

of variation between males and females. As the results of this study appear to support the
theoretical rationale and arguments already made concerning the impact of contextually-
related characteristics on cognitive ability performance, the present section will instead
focus brieﬂy on what the signiﬁcance of the present ﬁndings suggest for research and
practice aimed at minimizing gender performance discrepancies in future testing
instances.

As predicted by Bandura (1977) and later formulations by Gist and Mitchell
(1992), mechanical interests and experiences were strong predictors of an individual’s
reported level of mechanical self-efﬁcacy. Furthermore, the predictions drawn from
Bandura’s (1986) social cognitive theory of performance were also supported by the
ﬁnding that mechanical self-efﬁcacy was predictive of performance on the mechanical
ability test. These linked relationships supported the proposed mediation hypothesis,
indicating that both mechanical interests and mechanical experiences were predictive of
performance on the BMCT, though largely through their positive effects on mechanical
self-efﬁcacy. Based on these results, at least two conclusions of practical signiﬁcance
can be drawn. First, the original creators of the BMCT speciﬁcally designed the test so
as not to favor individuals who had previous training in physics or other related
specialties—operationalized here as mechanical experience—but instead to capture one’s
aptitude for perceiving and identifying physical principles in everyday life occurrences
(Bennett, 1969). Although the regression analyses presented for Hypothesis 2 would
appear to support this a priori presumption (via the observed non-signiﬁcant beta
coefﬁcient for mechanical experiences), there is strong evidence that the simultaneous

inclusion of mechanical interest in the regression equation acted as a suppressor. As

114

Cohen et al. (2003) suggest, a suppressor effect can often be identiﬁed by examining the
direction of the relationship in a simple correlation table with the observed beta weights
from regression; if a change in sigl (+/-) is observed, it is likely that the effects of the
variable are being masked by another included predictor. As this was indeed the case
found in the current results (cf. Table 10 and 13/18), the observed zero-order correlations
should be a better indicator of the actual relationship between mechanical experience and
BMCT performance. Thus, the results of this study do not appear to support the
presumption that mechanical experiences are not indicative of performance on the BMCT.
Although care should be taken in inferring such strict causal inferences from
correlational research, this the predictive relationship between mechanical experience and
mechanical ability test scores comes as no great surprise. If one considers speciﬁc
cognitive ability a form of specialized intelligence (which is not a signiﬁcant conceptual
leap given the theoretical similarity shared between g, s, and measures of intelligence,
Spearrnan, 1927; Vernon, 1950) there is a growing body of research spearheaded by
Stemberg (e. g., 1998; 1999; 2001; 2005) and colleagues (e.g., Stemberg, Grigorenko,
Ferrari, 2002) that suggests the same processes/techniques used to develop expertise are
applicable to developing intelligence. While the literature base on expertise is far too
large and extensive to describe in much detail here, empirical studies have generally
supported the notion that continued experience and deliberate practice within a given
domain greatly facilitates the acquisition of the skills and abilities necessary to reach a
high degree of proﬁciency related to that area of functioning (cf. Ericsson, Charness,
Feltovich & Hofﬁnan, 2006). Thus, despite what the original test publishers may have

posited, individuals seeking to improve their mechanical comprehension and subsequent

115

performance on related aptitude tests would seemingly be well served to tackle minor
household maintenance issues, tinker with common repair jobs, brieﬂy examine how a
particular mechanical object functions, and take advantage of the myriad other similar
experience-building opportunities in between. Furthermore, educators and vocational
specialists seeking to help students or clients excel on mechanical ability tests might
consider adding more practical demonstrations and learning tools to their curriculums to
encourage expertise development and familiarity with the common physical principles
included in the test materials.

The second conclusion of signiﬁcance that can be drawn from the present results
concerns the speciﬁc role mechanical interests and mechanical self-efﬁcacy play in
relation to mechanical comprehension test performance. Although self-efﬁcacy did
mediate the relationship between interests and experiences with performance, the results
suggested the relationship was only one of partial mediation—meaning both mechanical
background variables were still somewhat predictive of scores on the BMCT above and
beyond one’s reported level of mechanical self-efﬁcacy. This implies two additional
points of leverage for individuals or educators seeking to improve performance on the
mechanical ability test. First, because higher self-efﬁcacy scores were correlated
reasonably strongly with scores on the BMCT, improving an individual’s conﬁdence in
his/her ability to perform mechanical activities may result in better overall performance.
As presented previously, Bandura (1977; 1986) and Gist and Mitchell (1992) present a
useful theoretical framework from which strategies for improving self-efﬁcacy may be

selected and pursued further.

116

It should be noted that there is emerging empirical evidence to suggest there are
limits on the degree to which an individual’s self-efﬁcacy can be meaningfully enhanced
and whether self-efﬁcacy is even predictive of improved performance across multiple
measurement trials (e. g., Vancouver, Thompson, & Williams, 2001; Vancouver,
Thompson, & Tischner, 2002; Vancouver & Kendall, 2006). While the present study did
not adopt a within-person design and thus can not directly support or deny these
arguments, these results do not seem particularly troublesome given the typical
administration of ability tests such as mechanical comprehension. In large part, the
BMCT is used as a one-time, applicant or employee screening tool for a variety of
purposes (Super & Crites, 1962); as such, it would be somewhat unusual for an individual
to take the test more than once in close succession. Thus, the initial, “between-person”
level of mechanical self-efﬁcacy one possesses is likely still an important inﬂuencer of
subsequent performance on a mechanical comprehension test, and any effort to improve
those beliefs prior to taking the test should still be beneﬁcial.

The ﬁnal point of leverage for improving BMCT performance implied by the
partially mediated mechanical interests-mechanical comprehension relationship is not a
particularly insightful one, but important nonetheless: while acquiring more domain-
related interests should correlate with improved self-efﬁcacy, it also appears to inﬂuence
performance above and beyond the effects of these efﬁcacy beliefs. In this sense,
creating genuine interest in mechanically-related subject matter for an individual clearly
indicates a highly effective source for improving mechanical comprehension scores.
Unfortunately, according to many published accounts, interests are not a particularly

malleable individual difference. For example, Hall (2002) reports that by age 21 a

117

person’s interests largely stabilize; in fact, it is not uncommon to ﬁnd test-retest
correlations as high as .70 on many interest inventories, even over a period of 20 years.
Nevertheless, to the extent one could generate interest in mechanical activities and the
like (see Hidi & Baird, 1988, for a brief review of popular “interest development”
strategies and an example of one such attempt at generating interest in expository
reading), the beneﬁts to performance on a mechanical ability test such as the BMCT

appear to be great.

Capturing Male-Female Differences: An Investment Well-Spent

Consistent with previous research (e. g., Bennett & Cruikshank, 1942; Bennett,
1969; Maccoby & Jacklin, 1974; Anastasi, 1981, 1988; Antill & Cunningham, 1982;
Feingold, 1988; Stanley etal., 1992; Stumpf, 1995; Halpem, 2000), signiﬁcant and large
performance differences between males and females were observed in mechanical
comprehension. Males again outperformed females on the administered mechanical
ability test, on average scoring more than one standard deviation better than their female
counterparts. However and perhaps more interesting from a conceptual viewpoint,
meaningﬁll variation across sex was also found for all of the measured psychological
variables captured in the present experiment (see Table 11). This observation is
particularly meaningful given the oft-adopted approach in the literature whereby gender
differences in cognitive ability are so often simply summarized in the traditional,
categorical “participant sex” variable with no consideration as to why this might be the
case.

At the risk of crossing from theoretical discourse into facetious commentary, what

is so unfortunate about this empirical practice is that despite every breakthrough in

118

bioengineering, genetics and surgical procedure, biological sex is one of the few things
that a person can never fully rid themselves of or change for good. Thus, what good do
psychologists, as scientists and practitioners of human behavior, achieve by continuously
identifying those areas of functioning that exhibit male or female superiority without
further examining what makes a man or a woman better/worse at something? Armed
with that knowledge, one is simply able to make an observational statement. As the
model employed by the current study demonstrates, though, there appears to be relatively
strong evidence to suggest that the noted sex differences in mechanical ability test
performance can be attributed to differences in interests, experiences, and self-efﬁcacy
beliefs, which are subsequently inﬂuenced by gender role orientation and identiﬁcation.
Armed with this knowledge, future research could attempt to devise experiments that
manipulate or otherwise alter these variables in some meaningful fashion to determine
whether strategies or methodologies can be constructed that effectively and substantially
reduce mechanical ability differences. In this manner, observation moves into
explanation, explanation into action, and action hopeﬁrlly into useful implementation,
which should be the ultimate goal for any stream of scientiﬁc research.

If empirical inquiry does not employ such speciﬁc variables or make attempts to
investigate such questions with theoretically sound models, gender differences research
can easily digress into providing trivial “because he’s a man” or “because she’s a
woman” explanations for male-female discrepancies. Clearly the functional model of
gender differences around which the present research was organized did not completely
describe why males outperform females on tests of mechanical ability. However, that

was not the expectation for the model, nor should it have been. As the brief review of

119

biological explanations for differences in mechanical ability pointed out, there is deﬁnite
reason to believe that men and women are hard-wired to be more or less capable of
performing across various domains. What this model does represent, though, is an effort
to describe male-female differences in mechanical ability with individual differences in
psychological functioning rather than “participant sex”—with the hope being that future

research would beneﬁt from the identiﬁcation of potentially more manipulable variables.

Limitations

Like any empirical investigation, there are several limitations that should be taken
into account when considering the results of the present study. The ﬁrst such restriction,
which has already been alluded to in earlier discussions, concerns the correlational nature
of the research design. Importantly, as there were no direct manipulations applied to the
testing scenario or to the study’s participants, there is no evidence to support causal order
in the proposed gender differences model. Although theoretical rationale and previous
research in the broader literature were used to inform the implied direction of inﬂuence 1
between study variables, without experimental control the results of this study can only
hint at causation, not conﬁrm it.

An additional possible limitation concerns the use of college student as the
primary sample for this study. Aside ﬁom potential concerns of realism and external
validity, the research presented earlier on changing perceptions of gender appropriate
roles (e. g., Twenge, 1997; Holt & Ellis, 1998; Ritter, 2004; Duehr & Bono, 2006) implies
potential generational differences in the manner by which individuals would respond to
the gender-related measures of this study. For example, Table 11 indicates that the

degree respondents tended to not strongly endorse stereotypes about women or men,

120

regardless of their personal sex. However, this may be a symptom of the increasing
“androgynization” of more recent generations—wherein both men and women no longer
see great differences between the sexes (Twenge, 1997)—though a higher degree of
baseline endorsement might be expected in the older, more “traditional” working
population. Although it is unclear whether this conjecture has any real basis without
further sampling from a more diverse population, it is noted as a limitation of the present
study.

The remaining two limitations deal with the measurement of the study’s primary
constructs. As suggested in the test manual that accompanied the BMCT (Bennett, 1969),
the present experiment administered the mechanical ability test as a timed, 30-minute
exam. This protocol was followed to allow for an easy comparison of the overall .
performance of the present sample with that of the normative performance standards
gathered for the BMCT. In hindsight, though, this was likely an unnecessary aspect of
the testing situation given the purpose of the research at hand. The time limit, while
possibly increasing the realism of the testing context, could have been problematic had
the time pressure somehow differentially affected male and female test takers (or those
high or low on one of the individual difference variables of interest) by forcing those
individuals to rush through the test, change their test-taking strategy, encourage guessing,
etc. While there is no way to know whether this was the case in the present study, it is an
easily remedied problem. Thus in future research in which the purpose of administering
a cognitive ability test is simply to determine an individual’s absolute level of aptitude,

any time restriction unrelated to the desired research question should be removed so as to

121

prevent the possibility of introducing an unnecessary confound into the testing
environment.

Lastly, because all data used in this study were gathered with multiple-choice,
self-report questionnaires and testing instruments, mono-method biases or common
method variance may have been a concern as well (Podsakoff, MacKenzie, Lee, &
Podsakoff, 2003). However, as Spector (2006) argues, common method variance is not
invariably applicable simply because a single method of data collection was implemented
over the course of a study. Instead, one is better served by considering the nature of the
constructs being measured and any potential alternative, underlying explanations that
could have systematically altered participant responding to those measures. As such,
given that that the present study employed two classes of relatively similar variables (i.e.,
gender-related characteristics and contextually-related characteristics) response patterns
reﬂecting social desirability or a consistency motif would have been the most likely
causes of common method variance. Both of these factors generally result in inﬂated
correlations among the similar classes of variables, which could potentially bias the
results of subsequent hypothesis tests. However, aside from the relatively high
correlations observed between mechanical interests, experiences and self-efﬁcacy, the
interscale correlations among the measured variables were in the small to moderate range;
thus, if common method variance was an issue, it likely did not contribute signiﬁcantly to
the response patterns of participants. Nevertheless, future research should consider using
additional data sources (i.e., behavioral data, peer reports, etc.) to avoid any potential

problems associated with mono-method biases.

122

Conclusion

Although the acceptance of women into traditionally male dominated lines of
work has advanced considerably since Frances Gage addressed the AmericanEqual
Rights Association nearly a century and a half ago (Stanton, Anthony, & Gage, 1882), the
shift continues to be a slow process (Blau & Hendricks, 1979; England, 1981). The
passage of Equal Employment Opportunity laws and federal contract programs has
helped to speed the transition, but a large number of occupations continue to show
consistent and signiﬁcant inequalities in terms of the overall gender distributions of its
employees (Beller, 1982). The purpose of the present research was to test a functional
model of psychological antecedents proposed to predict the disparity in performance
observed between males and females on tests of mechanical comprehension, a factor
which may signiﬁcantly contribute to the noted occupational segregation trend in
vocations that require tests of mechanical ability for entry. Although not all the included
relationships hypothesized in the model were supported, a number of intriguing results
were discovered that offer incremental contributions to the literature’s current knowledge
base on gender differences in cognitive ability performance. Nonetheless, continued
efforts are needed to uncover further categories of individual difference variables that
explain why males and females appear to exhibit superiority across various domains of
psychological ﬁmctioning, in the hopes that we may one day fully understand the

processes through which such differences arise and live on through generations of human

beings.

123

APPENDIX A

Demographics/Background

Instructions:

 

Please respond to the following questions to the best of your ability.

1.

2.

What is your age?

What is your gender?
A. Male
B. Female

Are you of Hispanic origin?
A. Yes
B. No

What ethnicity do you consider yourself to be?

American Indian/Alaska Native

Asian

Black or Aﬁican American

Native Hawaiian and Other Paciﬁc Islander

White

American Indian/Alaska Native and White

Asian and White

Black or African American and White

American Indian/Alaska Native and Black or Aﬁican American

raoawpowr

What was your mother’s primary occupation when you were growing up?

 

Did your mother work primarily outside of the home?
A. Yes
B. No

What was your father’s primary occupation when you were growing up?

 

Did your father work primarily outside of the home?
A. Yes
B. No

124

APPENDIX B

Mechanical Interests, Knowledge and Experiences
(Adapted from Rechenberg, 2000)

Instructions:
Using the 5-point scale presented below, please respond to the following items to the best
of your ability.

 

 

 

1 2 3 4 5
Strongly l Disagree Neither agree Agree Strongly agree
disagree or disagree

Interests

1. I ﬁnd it rewarding to create or ﬁx something with my hands.

2. I enjoy reading magazines or watching television shows about mechanically related
topics (e. g., automobiles, new technologies, gadgets, etc.).

3. I enj0y learning about new techniques for performing mechanical or technical
activities (e. g., tips on home improvement/repair, auto maintenance, etc.).

4. When faced with an object that isn’t working properly (such as an appliance or a
bicycle), I enjoy trying to ﬁgure out the causes of the malfunction.

5. 1 like trying to discover how mechanical devices work (through observation, taking
things apart, etc.).

6. I am less interested in knowing how a mechanical device ﬁlnctions as I am in
knowing how to use the device. (R)

Knowledge

1. I am often asked to show or explain to others how to operate a piece of mechanical
equipment (e. g., run a lawnmower, use a power tool, use a sewing machine, etc.).

2. I have had much INFORMAL training related to mechanical or technical activities
(e. g., being taught how to ﬁx a car, leaky faucet, etc. by a relative or ﬁiend).

3. I have had much FORMAL training related to mechanical or technical activities (e.g.,
a school physics course, vocational/on-the-job training, etc).

125

When attempting to repair or assemble an unfamiliar object, I often can not ﬁgure out
what to do without referring to an instruction manual or help guide. (R)

I can usually ﬁgure out what is wrong with an object that is not working correctly.

I try to develop strategies or techniques (e. g. trial and error, working backwards, etc.)
for approaching mechanical or technical activities.

Experiences

1.

2.

When I was growing up, I often helped ﬁx things around the house.

I typically make repairs around the house when they are needed by myself rather than
ask for help.

1 have frequently taken part in skilled manual activities (e. g., home improvement,
auto repair, sewing projects).

I rarely take on do-it-yourself projects that require me to put together or build
something (e. g., bicycle, shelf, desk, etc.). (R)

I have performed many tasks that require the use of hand tools.

I have had many experiences where my mechanical abilities or skills were helpful in
ﬁxing a problem (e.g., changing a ﬂat tire, repairing a broken door, etc.).

126

APPENDIX C

Mechanical Self-Efﬁcacy
Instructions:
Please indicate how conﬁdent you are in your ability to successfully complete each of the
following tasks along the 7-point scale provided below.

 

1 2 3 4 5 6 7
Not at all Not very Very Completely
conﬁdent confident conﬁdent confident

1. Figure out how a mechanical item works (e. g., a ﬂashlight, simple engine, etc.) by
observing how its internal components operate (gears, belts, switches, etc.).

2. Explain why an object moved or acted a certain way based on what is happening in its
surrounding environment.

3. Identify the simple forces of physics that caused an object to move or behave in a
particular manner (e. g., momentum, gravity, centripetal force, etc.).

4. Predict what will happen in a situation when a physical element of the environment
changes (for example, predicting which way a tube of lipstick will roll along the ﬂoor
of your car based on the direction you turn).

5. Determine why an object (e. g., bicycle, mechanical clock, etc.) is no longer working
correctly.

6. Identify how the basic principles of physics (e. g., ﬁiction, pressure) allow a
mechanical item to operate.

7. Break down and identify the basic components (i.e. levers, pulleys, screws, etc.)
contained within a more complex machine.

8. Recognize basic principles of physics (6. g. forces of motion, gravity, etc.) in everyday
life.

127

APPENDIX D

Gender Role Identiﬁcation
Bern Sex Role Inventory (Bern, 1974)

Instructions:
Please read each of the characteristics listed and indicate to what degree each word or
phrase accurately describes yourself on the 7-point scale depicted below.

 

 

 

 

 

 

 

 

 

l 2 3 4 5 6 '7
Never true Almost Sometimes Can be not Sometimes Almost Always true
of me never true not true of true or true true of me always true of me
of me me of me of me
Masculine Items Feminine Items Neutral Items

49. Acts as a leader 11. Affectionate 51. Adaptable
46. Aggressive 5. Cheerful 36. Conceited
58. Ambitious 50. Childlike 9. Conscientious
22. Analytical 32. Compassionate 60. Conventional
l3. Assertive 53. Does not use harsh language 45. Friendly
10. Athletic 35. Eager to soothe hurt feelings 15. Happy
55. Competitive 20. Feminine 3. Helpful

4. Defends own beliefs l4. Easily ﬂattered 48. Inefﬁcient
37. Dominant 59. Gentle 24. Jealous
l9. Forceful 47. Gullible 39. Likable
25. Has leadership abilities 56. Loves children 6. Moody

7. Independent 17. Loyal 21. Reliable
52. Individualistic 26. Sensitive to the needs of others 30. Secretive
31. Makes decisions easily 8. Shy 33. Sincere
40. Masculine 38. Soft Spoken 42. Solemn

l . Self-reliant 23. Sympathetic 57. Tactﬁrl
34. Self—sufﬁcient 44. Tender 12. Theatrical
16. Strong personality 29. Understanding 27. Truthful
43. Willing to take a stand 41. Warm l8. Unpredictable
28. Willing to take risks 2. Yielding 54. Unsystematic

 

 

 

Note. The number preceding each item reﬂects the position of each adjective as it actually appears on the
survey.

128

 

APPENDIX E

Gender Stereotype Endorsement
(Adapted from Levy et al., 1998 and Blanton et al., 2002)

Instructions:

Some stereotypes are true. For example, men are stereotyped as physically stronger than
women. And indeed, a number of studies have supported the view of men as physically
stronger than women in general.

Listed below are some common gender stereotypes. As you read these, you may feel that
some are based on gender differences that really do exist or you may feel that the
stereotype has no basis in fact. Read each of the following stereotypes and rate the
degree to which you personally believe the stereotype is based on true gender differences
using the 5-point scale below. Please answer each question openly and honestly.

 

 

 

1 2 3 4 5
Not at all true A grain of Moderately Mostly true Absolutely
truth true true
I believe that...

1. Women have better organization skills than men.

2. Men have better spatial skills than women.

3. Women have greater verbal ability than men.

4. Men have better math skills than women.

5. Women have better interpersonal skills than men.

6. Men have better mechanical reasoning skills than women.

7. Women have better clerical skills (i.e., are good at typing quickly and accurately,
identifying mistakes in printed documents, etc.) than men.

8. Men have more overall intellectual ability than women.
9. ’Men have better analytic reasoning skills than women.

10. Women have more creative ability than men.

129

APPENDIX F

Informed Consent
Please read the information below completely and carefully:

This is a two-part study. For the ﬁrst portion of this study, we will be asking you to respond to an
online questionnaire asking about your interest, knowledge, experiences and conﬁdence level in
dealing with mechanically related content, as well as questions about your personality and beliefs.
You will also be asked to respond to a small number of demographic items that will help us
describe our research sample. We expect that it will take about 30 minutes to complete this ﬁrst
part of the study. For your participation in the ﬁrst part of the study, you will receive 1 point of
subject pool credit.

Following the online portion, you will be asked to come to the testing site and complete a test of
mechanical comprehension, which should last approximately 45 minutes. Upon completion of
this ﬁnal half of the study, you will receive an additional 2 points of subject pool credit. Thus,
fully participating in both halves of the study will be worth 3 points of subject pool credit (NOTE:
You may only participate in the second part of the study if you have previously completed the

ﬁrst part of the study).

There are no foreseeable risks associated with participating in this study. Your name and
information will remain conﬁdential. Your privacy will be protected to the maximum extent
allowable by law. The data will be saved for at least ﬁve years after it is collected and will only
be accessible by two faculty researchers and one graduate student. By typing your name below,
you indicate that you are free to refuse to participate in this project or any part of this project. You
may refuse to participate in certain procedures or answer certain questions. Your participation is
completely voluntary. You may choose not to participate at all and may discontinue your
participation at any time without penalty or loss of beneﬁts.

If you have any questions or concerns about your participation in this project, please contact
James Grand (by phone: (517) 355-2171, e-mail: grandjam@msu.edu, or by appointment: 348
Psychology Building). If you have questions or concerns regarding your rights as a study
participant, or are dissatisﬁed at any time with any aspect of this study, you may contact -
anonymously if you wish - Peter Vasilenko, Ph.D. Chair of the Human Research Protection
Program by phone: (517)355-2180, fax: (517)432-4503, email: irb@msu.edu, or regular mail:
202 Olds Hall, East Lansing, MI 48824.

Please mark the box that says "I agree to give my consent to participate" if you agree to
participate in this study. Mark "I do not want to participate" if you do not agree to participate in
this study. If you agree to participate, enter your name below and you will be taken to the survey
once this step is completed.

CI I agree to give my consent to participate
CI I do not want to participate .

First name Middle initial Last name

 

 

130

APPENDIX G

Participant Debrief
Participant Feedback/Debriefmg

Tests of mechanical ability are typically used as part of a selection process to choose applicants
for jobs that are mechanical in nature, such as assembly or maintenance workers. As these tests
can be used over a wide range of occupations that call for many different job-speciﬁc skills,
mechanical ability tests are designed to measure a person's general understanding of how the
principles of physics and motion work in our everyday lives.

Historically, males have signiﬁcantly outperformed females on such tests of mechanical
comprehension—however, the reasons for this large and consistent performance gap are unknown.
The purpose of the research study you have just participated in is to investigate the possible
causes and explanations for why gender differences in mechanical comprehension tests exist.
Aside from a person’s sex, the variables of interest in this study thought to inﬂuence mechanical
comprehension include:

1. Mil'lpnical interests, knowledge grid experiences

2. Mechﬂlical self-efﬁcacy — one’s conﬁdence level in their ability to adequately
perform a speciﬁc task

3. Gender Role Identiﬁcation — the degree to which an individual’s personality
characteristics are typically masculine or feminine

4. Gender Stereotype Endorsement - the degree to which an individual believes and
endorses the stereotypical differences between men and women

5. Social Penalties - the degree to which an individual has experienced sanctions or been
discouraged from pursuing activities related to mechanical ability

Prior to your participation in this study, neither you nor any of the research volunteers were
informed of the gender differences typically found in tests of mechanical comprehension. It was
necessary to withhold this information because of research ﬁndings which suggest that on tasks in
which a cultural stereotype exists (i.e., men have higher mechanical ability than women), any
information given to test takers that indicates that one group typically outperforms the other can
actually increase the performance gap between the two groups. Therefore, care was taken not to
present information on the exact purpose of the research study in order to prevent any undesirable
inﬂuences on overall test performance.

By better understanding why gender differences in mechanical comprehension testing exist, it
may be possible to develop more fair and accurate methods of testing to ensure that any
individual, regardless of sex, has equal opportunity to participate in jobs in which mechanical
ability is important. If you have further questions about the purpose of this study, how your
information will be used, or any other concerns, please contact:

James Grand

Phone: (517) 355-2171

E-mail: grandjam@msu.edu

Location: 346 Psychology Building, Michigan State University, East Lansing, MI 48824

131

APPENDIX H

Median Split Analyses for Hypotheses 6a-6d

In an attempt to more precisely replicate Antill and Cunningham’s (1992)
ﬁndings, 8 similar median split procedure and statistical analysis used by those
researchers was implemented. Table 22 presents the results of this procedure, which is
described in detail below.

To begin, a median split technique was used to divide participants in the current
database into the necessary gender roles. A random sample of 75 males and 75 females
were chosen from the dataset and used to calculate median scores for the masculinity and
femininity subscales of the BSR16. As Spence et al. (1975) suggest, high masculinity is
classiﬁed as any individual scoring above the median on the masculinity subscale and
high femininity as any individual above the median on the femininity subscale (and vice
versa for low masculinity and low femininity). Based on these distinctions, the median
scores were used as the cutoff point to form the four gender role groups—masculine,
feminine, androgynous, and undifferentiated—used by Antill and Cunningham (1982).
The masculine category was constructed to consist of high-masculine/low-feminine
individuals, the feminine category of low-masculine/high-feminine individuals, the
androgynous category of high-masculine/high-feminine individuals, and the

undifferentiated category of low-masculine/low—feminine individuals.

 

6 As Antill and Cunningham ( 1982) point out in Footnote 6 of their publication, equal numbers of males
and females were selected so that both sexes would equally contribute to the observed median scores.

132

Table 22
Mean BMC T Scores Based on Median-Split, BSRI Gender Role Identiﬁcation Categories
(H motheses 6a — 6d)

 

 

 

Category
M + A U + F t F + A U + M r M A U F
46.4 46.1 n. 46.9 n. 47.6 44.2 46.1
Males (58) (41) .181 44.7 (29) (70) 1.300 (37) (21) (33) 46.0 (8)
38.3 38.4 n. 38.4 38.3 n. 38.9 38.1 37.8 38.6
Females (72) (87) '031 (112) (47) '“2 (20) (52) (27) (60)
nsp = n.s.

Note. M = masculine; A = Androgynous; U = Undifferentiated; F = Feminine.
Numbers in parentheses indicate the number of subjects on which the mean next to it is
based.

The t column refers to the results of the t-test between the 2 means preceding it.

The second step performed by Antill and Cunningham was to nest males and
females within the gender role categories. Additionally, the authors also examined
differences in mechanical ability test performance within the masculine gender role by
creating separate high-masculine (masculine + androgynous individuals) and low-
masculine (undifferentiated + feminine individuals) groups for both males and females.
Although the researchers did not examine performance variations between high-feminine
(feminine + androgynous) and low-feminine (undifferentiated + masculine) individuals in
their original study, these comparison groups are necessary to examine the possibility of a
gender by gender role interaction. Thus in the present research, high-feminine and low-
feminine groups were formed for both males and females as well (beginning of Table 22).

The ﬁnal step in the analyses presented by Antill and Cunningham was to analyze
the observed mean differences between the comparison groups to determine if they were
signiﬁcant and in the predicted direction. As stated earlier, Antill and Cunningham did
not present results that tested for differences among the 2 gender (male, female) x 4

gender role (masculine, feminine, androgynous, undifferentiated) cell means, instead

133

focusing on within gender role effects. Thus, to obtain tests of the main effects and
interaction in the present study, a two-way AN OVA was conducted treating gender and
gender role identiﬁcation as between-subject factors. The results of this analysis indicate
that only the main effect for gender was predictive of performance differences in the
BMCT (F(1, 250) = 49.492, p < .001), with both the main effect of gender role
identiﬁcation (F(3, 250) = .734, n.s.) and the interaction effect (F(3, 250) = .363, n.s.)
failing to reach signiﬁcance].

In addition to examining differences among the gender by gender role subgroups,
differences within gender role were also examined. To do so, a series of t-tests were
conducted that speciﬁcally compared high-masculine males to low-masculine males and
high-feminine males to low-feminine males; the same analyses were then conducted for
females. However, no signiﬁcant differences were observed across any of the
comparisons indicating that variations in males’ or females’ reported strength of gender

role identiﬁcation did not affect their scores on the BMCT.

 

7 It is worth pointing out, however, that if only the absolute differences between cell means are examined
(as Antill and Cunningham, 1982, did), the current sample does reproduce the pattern of ﬁndings predicted
by Hypotheses 6a-6d Signiﬁcance testing aside, masculine males (47 .6) outperformed feminine males
(46.0), who outperformed masculine females (3 8.9), who outperformed feminine females (38.6). As the
two-way ANOVA revealed though, these differences were not signiﬁcant; it is the author’s belief that had
the same analysis been performed with Antill and Cunningham’s data, the same non-signiﬁcant result
would have been found among their observed cell means as well.

134

REFERENCES

Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting
interactions. Newbury Park, CA: Sage.

Anastasi, A. (1981). Sex differences: Historical perspectives and methodological
implications. Developmental Review, 1, 187-206.

Anastasi, A. (1988). Psychological testing (6th ed.). New York: Macmillan.

Antill, J. K., & Cunningham, J. D. (1982). Sex differences in performance on ability tests
as a function of masculinity, femininity, and androgyny. Journal of Personality &
Social Psychology, 42(4), 718-728.

Bandura, A. (1969). Social-learning theory of identiﬁcatory processes. In D. A. Goslin
(Ed.) Handbook of socialization theory and research (pp. 213-262). Chicago:
Rand McNally.

Bandura, A. (1977). Self efﬁcacy: Toward a unifying theory of behavioral change.
Psychological Review, 84(2), 191-215.

Bandura, A. (1986). Social foundations of thought and action. Englewood Cliffs, NJ:
Prentice-Hall.

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in
social psychological research: Conceptual, strategic, and statistical considerations.
Journal of Personality and Social Psychology, 51(6), 1 173-1 182.

Beere, C. A. (1990). Gender roles: A handbook of tests and measures. Westport, CT:
Greenwood Press.

Beller, A. H. (1982). Occupational segregation by sex: Determinants and changes. The
Journal of Human Resources, 1 7(3), 371-392.

Bern, S. L. (1974). The measurement of psychological androgyny. Journal of Consulting
& Clinical Psychology, 42(2), 155-162.

Bern, S. L. (1981). Gender schema theory: A cognitive account of sex-typing.
Psychological Review, 88, 354-364.

Bern, S. L. (1985). Androgyny and gender schema theory: A conceptual and empirical
integration. In T. B. Sonderegger (Ed.) Nebraska Symposium on Motivation:
Psychology of Gender. Lincoln, NB: University of Nebraska Press.

135

Benett, G. K., & Cruikshank, R. (1942). Sex differences in the understanding of
mechanical problems. Journal of Applied Psychology, 26(2), 121-127.

Bennett, G. K. (1969). Manual for the Bennett mechanical comprehension test, Forms S
& T. New York: The Psychological Corporation.

Bennett, G. K. (2006). Bennett mechanical comprehension test, Form S. San Antonio,
TX: Harcourt Assessment Incorporated.

Bernard, M. E., Boyle, G. J ., & Jackling, I. (1990). Sex-role identity and mental ability.
Personality and Individual Dlﬂerences, I I , 213-217.

Blanton, H., Christie, C., & Dye, M. (2002). Social identity versus reference frame
comparisons: The moderating role of stereotype endorsement. Journal of
Experimental Social Psychology, 38(3), 253-267.

Blau, F. D., & Hendricks, W. E. (1979). Occupational segregation by sex: Trends and
prospects. The Journal of Human Resources, 14(2), 197-210.

Brown, S. M. (1979). Male versus female leaders: A comparison of empirical studies. Sex
Roles, 5(5), 595-611.

Cantoni, L. P. (1955). High school tests and measurements as predictors of occupational
status. Journal of Applied Psychology, 39(4), 253-255.

Chen, G., Gully, S. M., Whiteman, J. A., & Kilcullen, R. N. (2000). Examination of
relationships among trait-like individual differences, state-like individual
differences, and learning performance. Journal of Applied Psychology, 85(6),
83 5-847 .

Chen, G., Gully, S. M., & Eden, D. (2001). Validation of a new general self-efﬁcacy
scale. Organizational Research Methods, 4(1), 62-83.

Choi, N., & Fuqua, D. R. (2003). The structure of the Bern Sex Role Inventory: A

summary of 23 validation studies. Educational and Psychological Measurement,
63(5), 872-887.

Cleveland, J ., Stockdale, M. S., & Murphy, K. R. (2000). Women and men in

organizations: Sex and gender issues at work. Mahwah, N. J .: Lawrence Erlbaum
Associates.

Cohen, J. (1983). The cost of dichotomization. Applied Psychological Measurement, 7,
249-253.

Cohen, J ., Cohen, R, West, S. G., & Aiken, L. S. (2003). Applied multiple
regression/correlation analysis for the behavioral sciences (3rd ed.). Mahwah,
New Jersey: Lawrence Erlbaum Associates.

136

Constantinople, A. (1973). Masculinity-femininity: An exception to a famous dictum?
Psychological Bulletin, 80(5), 389-407.

Cox, J. W. (1928). Mechanical aptitude: Its existence, nature and measurement. London:
Methuen & Co. Ltd.

Devine, P. G. (1989). Stereotypes and prejudice: Their automatic and controlled
components. Journal of Personality and Social Psychology, 56(1), 5-18.

Eagly, A. H., J ohannesen-Schmidt, M. C., & van Engen, M. L. (2003). Transformational,
transactional, and laissez-faire leadership styles: A meta-analysis comparing
women and men. Psychological Bulletin, 129(4), 569-591. '

Eagly, A. H., & Karau, S. J. (2002). Role congruity theory of prejudice toward female
leaders. Psychological Review, 109(3). 573-598.

Eccles, J. S. (1987). Gender roles and women's achievement-related decisions.
Psychology of Women Quarterly, 11(2), 135-172.

England, P. (1981). Assessing trends in occupational sex segregation, 1900-1976. In I.
Berg (Ed.), Sociological perspectives on labor markets (pp. 273-295). New York:
Academic Press.

Ericsson, K. A., Charness, N., F eltovich, P. J ., Hoffman, R. R. (Eds.) (2006). The
Cambridge handbook of expertise and expert performance. New York:
Cambridge University Press.

F eingold, A. (1988). Cognitive gender differences are disappearing. American
Psychologist, 43(2), 95-103.

Flynn, J. R. (1984). IQ gains and the Binet decrements. Journal of Educational
Measurement, 21(3), 283-290.

Geschwind, N., & Galaburda, A. M. (1987). Cerebral lateralization: Biological
mechanisms, associations, and pathology. Cambridge, MA: MIT Press.

Ghiselli, E. E. ,& Brown, C. W. (1951). Validity of tests for auto mechanics. Journal of
Applied Psychology, 35(1), 23- 24.

Gist, M. E., & Mitchell, T. R. (1992). Self-efﬁcacy: A theoretical analysis of its

determinants and malleability. The Academy of Management Review, 1 7(2), 183-
21 1.

137

Guertin, A. S., Guertin, W. H., & Ware, W. B. (1981). Distortion as a function of the
number of factors rotated under varying levels of common variance and error.
Educational and Psychological Measurement, 41(1), 1-9.

Guilford, J. P. (1947). The discovery of aptitude and achievement variables. Science,
106(2752), 279-282.

Guilford, J. P. (1948). Factor analysis in a test-development program. Psychological
Review, 55(2), 79-94.

Guilford, J. P., & Lacey, J. I. (Eds). (1947). Printed classiﬁcation tests (AAF Aviation
Psychology Program, Research Reports, Rep. No. 5). Washington, DC: US.
Government Printing Ofﬁce.

Hall, D. T. (2002). Careers in and out of organizations. Thousand Oaks, CA: Sage
Publications.

Halpem, D. F. (2000). Sex diﬂerences in cognitive abilities (3rd ed.). Mahwah, NJ: L.
Erlbaum Associates.

Hamburg, D. A., & Lunde, D. T. (1966). Sex hormones in the development of sex
differences in human behavior. In E. E. Maccoby (Ed.) The Development of Sex
Diﬂerences (pp. 1-24). Stanford, CA: Stanford University Press.

Hamilton, C. J. (1995). Beyond sex differences in visuo-spatial processing: The impact of
gender trait possession. British Journal of Psychology, 86, 1-20.

Harrell, W., & Faubion, R. (1940). Selection tests for aviation mechanics. Journal of
Consulting Psychology, 4(3), 104-105.

Harris, J. (1978). Sex differences in spatial ability: Possible environmental, genetic, and
neurological factors. In M. Kinsboume (Ed.), Asymmetrical function of the brain.
Cambridge, NY: Cambridge University Press.

Harris, J. (1981). Sex-related variations in spatial skill. In L. S. Liben, A. H. Patterson &
N. Newcombe (Eds), Spatial representation and behavior across the life span:
Theory and application (pp. 83-125). New York: Academic Press.

Hidi, S., & Baird, W. (1988). Strategies for increasing text-based interest and students’
recall of expository texts. Reading Research Quarterly, 24(4), 465-483.

Holt, C. L., & Ellis, J. B. (1998). Assessing the current validity of the Bern Sex Role
Inventory. Sex Roles, 39, 929-941.

Hunter, J. E., & Schmidt, F. L. (1990). Dichotomization of continuous variables: The
implications for meta-analysis. Journal of Applied Psychology, 75, 334-349.

138

Huston, A. C. (1983). Sex-typing. In E. M. Hetherington (Ed.), Handbook of child
psychology: Vol. 4. Socialization, personality, and social development (pp. 387-
467). New York: Wiley.

J acklin, C. N., & Reynolds, C. (1993). Gender and childhood socialization. In A. Beall &
R. J. Stemberg (Eds). The Psychology of Gender (pp. 197-214). New York:
Guilford Press.

Judge, T. A., & Bono, J. E. (2001). Relationship of core self-evaluations traits-self-
esteem, generalized self-efﬁcacy, locus of control, and emotional stability-with

job satisfaction and job performance: A meta-analysis. Journal of Applied
Psychology, 86(1), 80-92.

Kenny, DA. (2008). Mediation. Retrieved November .27, 2008 from
http://davidakenny.net/cm/mediate.htm#BK.

Klare, G. R., Gustafson, L. M., Mabry, J. E., & Shuford, E. H. (1955). The relationship of
immediate retention of technical training material to career preferences and
aptitudes. Journal of Educational Psychology, 46(6), 321-329.

Lane, J ., & Lane, A. (2001). Self-efﬁcacy and academic performance. Social Behavior &
Personality: An International Journal, 29(7), 687-694.

Lepore, L., & Brown, R. (2000). Category and stereotype activation: Is prejudice
inevitable? In C. Stangor (Ed.), Stereotypes and prejudice (pp. 119-137).
Philadelphia, PA: Psychology Press.

Levonian, E., & Comrey, A. L. (1966). Factorial stability as a function of the number of
orthogonally-rotated factors. Behavioral Science, 11(5), 400-404.

Levy , S. R., Stroessner, S. J ., & Dweck, C. S. (1998). Stereotype formation and
endorsement: The role of implicit theories. Journal of Personality and Social
Psychology, 74(6), 1421-1436.

Lott, B., & Maluso, D. (1993). The social learning of gender. In A. Beall & R. J.

Stemberg (Eds). The Psychology of Gender (pp. 99-123). New York: Guilford
Press.

Maccoby, E. E., & Jacklin, C. N. (1974). The psychology of sex differences. Stanford, CA:
Stanford University Press.

MacKinnon, D.P., Fairchild, A.J., Fritz, MS. (2007). Mediation analysis. Annual Review
of Psychology, 58, 593-614.

Massa, L. J ., Mayer, R. E., Bohon, L. M. (2005). Individual differences in gender role
beliefs inﬂuence spatial ability test performance. Learning and Individual
Differences, 15(2), 99-111.

139

Martin, G. C. (1951). Test batteries for trainees in auto mechanics and apparel design.
Journal of Applied Psychology, 35(1), 20-22.

Mathieu, J. E., Martineau, J. W., & Tannenbaum, S. I. (1993). Individual and situational
inﬂuences on the development of self-efﬁcacy: Implications for training
effectiveness. Personnel Psychology, 46(1), 125-147.

Maxwell, S. E., & Delaney, H. D. (1993). Bivariate median splits and spurious statistical
signiﬁcance. Psychological Bulletin, 1 13(1), 181-190.

Messent, P. R. (1976). Female hormones and behaviour. In B. Lloyd & J. Archer (Eds)
Exploring Sex Differences (pp. 185-212). London: Academic Press.

Milton, G. A. (1957). The effects of sex-role identiﬁcation upon problem-solving skill.
Journal of Abnormal and Social Psychology, 55(2), 208-212.

Moritz, S. E., Feltz, D. L., Fahrbach, K. R., & Mack, D. E. (2000). The relation of self-
efﬁcacy measures to sport performance: A meta-analytic review. Research
Quarterly for Exercise and Sport, 71(3), 280-294.

Muchinsky, P. M. (2004). Mechanical aptitude and spatial ability testing. In J. C. Thomas
(Ed.), Comprehensive handbook of psychological assessment, Vol. 4: Industrial
and organizational assessment (pp. 21-33). Hoboken, NJ: John Wiley & Sons, Inc.

Nash, S. C. (1975). The relationship among sex-role stereotyping, sex-role preference,
and the sex difference in spatial visualization. Sex Roles, 1(1), 15-32.

Nash, S. C. (1979). Sex role as a mediator of intellectual ﬁmctioning. In M. A. Wittig &
A. C. Petersen (Eds), Sex-related diﬁferences in cognitive functioning:
Developmental issues (pp. 263-302). New York: Academic Press.

Newcombe, N. & Dubas, J. S. (1992). A longitudinal study of predictors of spatial ability
in adolescent females. Child Development, 63, 37-46.

Nosek, B. A., Banaji, M. R., & Greenwald, A. G. (2002). Math = male, me = female,
therefore math 2% me. Journal of Personality & Social Psychology, 83(1), 44-59.

Patterson, C. H. (1956). The prediction of attrition in trade school courses. Journal of
Applied Psychology, 40(3), 154-158.

Patterson, D. G., Elliot, R. M., Anderson, L. D., Toops, H. A., & Heidbreder, E. ( 1930).
Minnesota mechanical ability tests. Minneapolis: University of Minnesota Press.

140

Phillips, J. M., & Gully, S. M. (1997). Role of goal orientation, ability, need for
achievement, and locus of control in the self-efﬁcacy and goal-setting process.
Journal of Applied Psychology, 82(5), 792-802.

Piedmont, R. L., & Weinstein, H. P. (1994). Predicting supervisor ratings of job
performance using the NEO Personality Inventory. Journal of Psychology, 128,
255-265.

Podsakoff, P. M., MacKenzie, S. B., Lee, J .-Y., Podsakoff, N. P. (2003). Common
method biases in behavioral research: A critical review of the literature and
recommended remedies. Journal of Applied Psychology, 88(5), 879-903.

Rechenberg, C. (2000). Understanding gender differences in mechanical performance.
Unpublished doctoral dissertation, University of Akron, Ohio.

Rogers, L. (1976). Male hormones and behaviour. In B. Lloyd & J. Archer (Eds)
Exploring Sex Differences (pp. 157-184). London: Academic Press.

Rummel, R. J. (1970). Applied factor analysis. Evanston, IL: Northwestern University
Press.

Schmader, T. (2002). Gender identiﬁcation moderates stereotype threat effects on

women’s math performance. Journal of Experimental Social Psychology, 38(2),
1 94-201 .

Schmader, T., Johns, M., & Barquissau, M. (2004). The costs of accepting gender
differences: The role of stereotype endorsement in women’s experience in the
math domain. Sex Roles, 5 0(1 1/12), 835-850.

Schmitt, N., & Stuits, D. M. (1985). Factors deﬁned by negatively keyed items: The

result of careless respondents? Applied Psychological Measurement, 9(4), 367-
373. ‘

Schwoerer, C. R., May, D. R., Hollensbe, E. C., & Mencl, J. (2005). General and speciﬁc
self-efﬁcacy in the context of a training intervention to enhance performance
expectancy. Human Resource Development Quarterly, 16(1), 1 1 1-129.

Sherer, M., Maddux, J. E., Mereandante, B., Prentice-Dunn, 8., Jacobs, B., & Rogers, R.

W. (1982). The Self-Efﬁcacy Scale: Construction and validation. Psychological
Reports, 51, 663-671.

Signorella, M. L., & Jamison, W. (1986). Masculinity, femininity, androgyny, and

cognitive performance: A meta-analysis. Psychological Bulletin, 100(2), 207-
228.

141

 

Signorella, M. L., & Vegaga, M. E. (1984). A note on gender stereotyping of research
topics. Personality and Social Psychology Bulletin, 10, 107-109.

Spearrnan, C. (1927). The abilities of man: Their nature and measurement. New York:
MacMillan.

Spector, P. E. (2006). Method variance in organizational research: Truth or urban legend?
Organizational Research Methods, 9(2), 221-232.

Spence, J. T., Helmreich, R., & Stapp, J. (1975). Ratings of self and peers on sex role
attributes and their relation to self-esteem and conceptions of masculinity and
femininity. Journal of Personality & Social Psychology, 32(1), 29-39.

Spencer, S. J ., Steele, C. M., & Quinn, D. M. (1999). Stereotype threat and women’s
math performance. Journal of Experimental Social Psychology, 35(1), 4-28.

Stajkovic, A. D., & Luthans, F. (1998). Self-efﬁcacy and work-related performance: A
meta-analysis. Psychological Bulletin, 124(2), 240-261 .

Stangor, C. (Ed.). (2000). Stereotypes and prejudice. Philadelphia, PA: Psychology Press.

Stangor, C., & Lange, J. (1994). Mental representations of social groups: Advances in
conceptualizing stereotypes and stereotyping. Advances in Experimental Social
Psychology, 26, 357-416.

Stanley, J. C., Benbow, C. P., Brody, L. E., Dauber, S., & Lupkowski, A. (1992). Gender
differences on eighty-six nationally standardized aptitude and achievement tests.
In N. Colangelo, S. G. Assouline, & D. L. Ambroson (Eds) Talent development,
Vol. 1: Proceedings from the 1991 Henry B. and Jocelyn Wallace National
Research Symposium on Talent Development (pp. 42-65). Unionville, NY:
Trillium Press.

Stanton, E. C., Anthony, S. B., & Gage, M. J. (Eds). (1882). History of woman suﬂrage.
New York: Fowler & Wells.

Steele, C. M., & Aronson, J. (1995). Stereotype threat and the intellectual test
performance of African Americans. Journal of Personality and Social Psychology,
69(5), 797-811.

Stein, A. H., & Bailey, M. M. (1973). The socialization of achievement orientation in
females. Psychological Bulletin, 80, 345-366.

Stenquist, J. L. (1923). Measurements of mechanical ability. New York: Teachers
College, Columbia University.

Stemberg, R. J. (1998). Metacognition, abilities, and developing expertise: What makes
an expert student? Instructional Science, 26(1/2), 127-140.

142

Stemberg, R. J. (1999). Intelligence as developing expertise. Contemporary Educational
Psychology, 24(4), 359-375.

Stemberg, R. J. (2001). Giftedness as developing expertise: A theory of the interface
between high abilities and achieved excellence. High Ability Studies, 12(2), 159-
179.

Stemberg, R. J. (2005). Intelligence, competence, and expertise. In A. J. Elliot & C. S.
Dweck (Eds) Handbook of competence and motivation (pp. 15-30). New York:
Guilford Publications.

Stemberg, R. J ., Grigorenko, E. L., & Ferrari, Michel (2002). Fostering intellectual
excellence through developing expertise. In M. Ferrari (Ed.) The pursuit of
excellence through education (pp. 57-83). Mahwah, NJ: Lawrence Erlbaum
Associates.

Stewart, G. L. (1999). Trait bandwidth and stages of job performance: Assessing
differential effects for conscientiousness and its subtraits. Journal of Applied
Psychology, 84, 959-968.

Stumpf, H. (1995). Gender differences in performance on tests of cognitive abilities:
Experimental design issues and empirical results. Learning and Individual
Diﬂerences, 7(4), 275-287.

Super, D. E., & Crites, J. O. (1962). Appraising vocational ﬁtness by means of
psychological tests (Rev. ed.). New York: Harper.

Sundet, J. M., Barlaug, D. G., & Torjussen, T. M. (2004). The end of the Flynn Effect? A
study of secular trends in mean intelligence test scores of Norwegian conscripts
during half a century. Intelligence, 32(4), 349-362.

Tajfel, H., & Turner, J. C. (1986). Social identify theory of intergroup behavior. In W.
Austin & S. Worchel (Eds), Psychology of intergroup relations (2“d ed., pp. 7-24).
Chicago: Nelson-Hall.

Teasdale, T. W., & Owen, D. R. (2000). Forty-year secular trends in cognitive abilities.
Intelligence, 28(2), 115-120.

Thurstone, L. L. (1948). Psychological implications of factor analysis. American
Psychologist, 3, 402-408.

Tifﬁn, J ., Knight, F., & Asher, E. (1946). The psychology of normal people. Boston: D.C.
Heath and Company.

Tinsley, H. E. A., & Tinsley, D. J. (1987). Uses of factor analysis in counseling
psychology research. Journal of Counseling Psychology, 34(4), 414-424.

143

Tipton, R. M., & Worthington, E. (1984). The measurement of generalized self-efﬁcacy:
A study of construct validity. Journal of Personality Assessment, 48, 545-548.

Twenge, J. M. (1997). Changes in masculine and feminine traits over time: A meta-
analysis. Sex Roles, 36(5/6), 305-325.

van den Berg, P. T., & Feij, J. A. (2003). Complex relationships among personality traits,
job characteristics, and work behaviors. International Journal of Selection and
Assessment, 11(4), 326-339.

van der Maas, H. L. J ., Dolan, C. V., Grasman, R. P. P. P., Wicherts, J. M., Huizenga, H.
M., & Raijmakers, M. E. J. (2006). A dynamical model of general intelligence:

The positive manifold of intelligence by mutualism. Psychological-Review,
113(4), 842-861.

van Engen, M. L., & Willemsen, T. M. (2004). Sex and leadership styles: A meta-
analysis of research published in the 1990s. Psychological Reports, 94(1), 3-18.

Vancouver, J. B., & Kendall, L. N. (2006). When self-efﬁcacy negatively relates to

motivation and performance in a learning context. Journal of Applied Psychology,
91(5), 1146-1153.

Vancouver, J. B. ”Thompson C. M., & Tischner, E. C. (2002). Two studies examining
the negative effect of self-efﬁcacy on performance. Journal of Applied
Psychology, 87(3), 506- 516.

Vancouver, J. 8., Thompson, C. M., & Williams, C. M. (2001). The changing signs in the
relationships among self-efﬁcacy, personal goals, and performance. Journal of
Applied Psychology, 86(4), 605-620.

Vernon, P. E. (1950). The structure of human abilities. London: Methuen.

Vinchur, A. J ., Schippmann, J. S., Switzer, F. S., & Roth, P. L. (1998). A meta-analytic
review of predictors of job performance for salespeople. Journal of Applied

Psychology, 83, 586-597.

Walton, G. M., & Cohen, G. L. (2003). Stereotype lift. Journal of Experimental Social
Psychology, 39(5), 456-467.

Weisen, J. P. (1999). Technical manual for the Weisen test of mechanical aptitude
( WT MA ). Newton, MA: Applied Personnel Research.

Williams, J. E., & Bennett, S. M. (1975). The deﬁnition of sex stereotypes via the
Adjective Check List. Sex Roles, 1(4), 327-337.

144

Williams, J. E., & Best, D. L. (1990). Measuring sex stereotypes: A multination study
(Rev. ed.). Newbury Park, CA: Sage.

Wittenbom, J. R. (1945). Mechanical ability, its nature and measurement, 1: An analysis
of the variables employed in the preliminary Minnesota experiment. Journal of
Educational and Psychological Measurement, 5, 243- 262.

Wolff, W. M., &North, A. J. (1951). Selection of mun1c1pal ﬁremen. Journal of Applied
Psychology, 35(1), 25-29.

Wood, W., & Eagly, A. H. (2002). A cross-cultural analysis of the behavior of women

and men: Implications for the origins of sex differences. Psychological Bulletin,
128(5), 699-727. ~

Woodruff, S. L., & Cashman, J. E. (1993). Task, domain, and general efﬁcacy: A
reexamination of the Self-Efficacy Scale. Psychological Reports, 72, 422-432.

Yeo, G. B., & Neal, A. (2006). An examination of the dynamic relationship between self-

efﬁcacy and performance across levels of analysis and levels of speciﬁcity.
Journal of Applied Psychology, 91 (5), 1088-1 101.

145